Bloomberg Pricing in

Document Sample
Bloomberg Pricing in Powered By Docstoc
					Pricing in the
IT World
Master’s Thesis in Computer Science

By Michael R. Werder

of Rüti ZH, Switzerland
Register No. 99-704-090

Reinhard Riedl, PhD
Department of Information Technology
University of Zurich, Switzerland

Submitted on January 16, 2004
Service orientation is a key paradigm in today’s IT world and manifests itself in both
organizational as well as technological developments. In order to be commercially applied,
evolving service-oriented technologies such as web services and grid computing require
feasible business models.

In this context, the thesis at hand is concerned with the pricing of IT-based services. In a
comprehensive approach, economic, organizational, managerial, and technological aspects of
service pricing are examined. Theoretical and practical models for service pricing are
discussed and enabling architectures for pricing in grid computing as well as in markets
organized through service brokers are presented.

Serviceorientierung ist ein dominierendes Paradigma in der heutigen IT Welt und manifestiert
sich sowohl in organisatorischen als auch in technologischen Entwicklungen. Um
kommerziell angewandt werden zu können, benötigen aufstrebende Technologien wie Web
Services oder Grid Computing praktikable Geschäftsmodelle.

Die vorliegende Arbeit befasst sich mit der Preisbildung für IT-basierte Dienste in diesem
Kontext.   In   einem    umfassenden    Ansatz     werden   ökonomische,   organisatorische,
betriebswirtschaftliche und technologische Aspekte der Preisbildung für Dienste untersucht.
Nebst der Diskussion theoretischer und praktischer Modelle der Preisbildung werden
Architekturen für die Preisbildung im Umfeld des Grid Computing sowie mit Hilfe von
Service Broker präsentiert.


I am grateful to the following people who have supported me in the writing of this thesis:

   •   Dr. Reinhard Riedl for giving me the opportunity to write this thesis under his

   •   Peter Weibel and Marc Lütolf for sharing their professional experience with me

   •   Christa Werder, Andrea Aepli and Martin Luchsinger for critically proofreading my
       work and ensuring that my morale was never fading

   •   My parents, Christa and Jürg, for feeding me and providing me with a safe haven
       during the last twenty-four years

   •   Andrea for her love, patience, and understanding

                                           Thank you.

Table of Contents

1   Introduction .................................................................................................................. 1

2   Terminology .................................................................................................................. 6
    2.1 Service................................................................................................................... 6
    2.2 IT – Information Technology................................................................................ 7
    2.3 Services in the IT Domain..................................................................................... 7
        2.3.1     E-Services............................................................................................... 7
         E-Services from the Marketing Perspective............................ 8
         Web Services........................................................................... 8
         Traditional Services vs. E-Services....................................... 11
        2.3.2     Intra-Corporate IT Services.................................................................. 12
        2.3.3     The IT Services Industry ...................................................................... 13
         Services Provided by Internet Service Providers .................. 14
         Services Provided by Application Service Providers............ 17
         Services Enabled by Grid Computing................................... 20
    2.4 Summary ............................................................................................................. 24

3   The Paradigm of Service-Oriented Computing from Various Perspectives......... 25
    3.1 Technological Perspective on the Paradigm of Service-Oriented Computing.... 25
         3.1.1  Service-Oriented Architecture (SOA).................................................. 25
         3.1.2  Mobile Agents and Peer-to-Peer Computing: Alternatives to the ...........
                Client-Server Model? ........................................................................... 27
       Mobile Agents ....................................................................... 27
       Peer-to-Peer Computing........................................................ 29
       Conclusion............................................................................. 31
    3.2 Managerial Perspective on the Paradigm of Service-Oriented Computing ........ 32
         3.2.1  Quality Management: Service Management Frameworks and SLAs .. 33
         3.2.2  Cost Management: Chargeback ........................................................... 34
         3.2.3  Management of IT Risk and Resilience ............................................... 35
         3.2.4  Management of Web Services.............................................................. 38
    3.3 Communicative Perspective on the Paradigm of Service-Oriented Computing . 39
         3.3.1  Common Areas of Dispute................................................................... 39
         3.3.2  Communicating the Value of IT........................................................... 41
         3.3.3  The IT Department: Cost or Profit Center?.......................................... 42
    3.4 Economic Perspective on the Paradigm of Service-Oriented Computing .......... 43
    3.5 Conclusion........................................................................................................... 46

4   An Overview of Service Pricing ................................................................................ 48
    4.1 The Basic Principles of Pricing........................................................................... 48
        4.1.1   Supply, Demand and Price ................................................................... 48
        4.1.2   Factors Affecting Price......................................................................... 50
        4.1.3   Pricing Policies..................................................................................... 53
        4.1.4   Common Approaches for Service Pricing............................................ 55
        4.1.5   Summary: A Framework to Grasp the Different Aspects of Pricing ... 61

5   The Specialties of Pricing IT-Based Services........................................................... 65
    5.1 Lock-in ................................................................................................................ 65
    5.2 Versioning ........................................................................................................... 65
    5.3 Network Effects................................................................................................... 66
    5.4 Increased Perceived Risk .................................................................................... 67
    5.5 Global Competition ............................................................................................. 67
    5.6 Further Considerations ........................................................................................ 68
         5.6.1    Pricing for SLA-Based Service Provision............................................ 68
         Appropriateness of SLAs Within Smaller Corporations....... 72
         5.6.2    Pricing the Services of Dynamic Virtual Enterprises........................... 73
         5.6.3    Pricing Service Reservations................................................................ 74
         5.6.4    Pricing Guaranteed Service Levels and Risk Assumption................... 76
         5.6.5    Pricing for Mid- and Long-Term Contracts ......................................... 77
         5.6.6    The Impact of Standards on Service Markets ...................................... 79
    5.7 Summary ............................................................................................................. 81

6   Pricing in Practice ...................................................................................................... 82
    6.1 The Market Participants’ Perspectives on Service Pricing ................................. 82
         6.1.1    From the Providers’ Perspective .......................................................... 82
         6.1.2    From the Brokers’ Perspective............................................................. 84
         6.1.3    From the Third Parties’ Perspective..................................................... 86
         6.1.4    From the Consumers’ Perspective........................................................ 88
    6.2 Real-World Restrictions for Theoretical Pricing Models and Methods.............. 90
         6.2.1    Causes of Problems .............................................................................. 90
         6.2.2    Requirements on IT-Infrastructure and Administration....................... 92
    6.3 Effects of Different Pricing Methods on the Market, and on Supplier and
         Consumer Behavior............................................................................................. 94
    6.4 Selecting Pricing Models .................................................................................... 96
         6.4.1    Success Criteria for Pricing Models..................................................... 96
         6.4.2    Choosing the Right Pricing Model....................................................... 98
         Example: Flat-Rate vs. Usage-Based Pricing ..................... 100
    6.5 Summary ........................................................................................................... 102

7   Service Pricing for ISPs, ASPs, and in Grid Computing...................................... 104
    7.1 Pricing the Services of ISPs .............................................................................. 104
         7.1.1    Overview of Theoretical Models........................................................ 104
         7.1.2    ISPs’ System Architecture ................................................................. 107
         7.1.3    ISP Pricing in Practice........................................................................ 107
         7.1.4    ISP Pricing Schemes Commonly Found in Practice .......................... 108
         7.1.5    Pricing Between ISPs: Peering........................................................... 110
    7.2 Pricing the Services of ASPs............................................................................. 112
         7.2.1    Overview of Theoretical Models........................................................ 112
         7.2.2    ASPs’ System Architecture................................................................ 114
         7.2.3    ASPs’ Cost Structure.......................................................................... 116
         7.2.4    ASP Pricing in Practice ...................................................................... 117
    7.3 Pricing of Services in Grid Computing ............................................................. 118
         7.3.1    Overview of Theoretical Models........................................................ 118
         7.3.2    Grid Service Providers’ Architecture................................................. 121
         7.3.3    Organizational Aspects ...................................................................... 122
         7.3.4    Differences between the Pricing of Data, Computation,
                  Communication, Applications and Knowledge in Grid Computing .. 124

             7.3.5   Pricing in Scientific Grids vs. Pricing in Commercial Grids ............. 127
     7.4     Summary: Differences between Service Pricing for ISPs, ASPs and in ................
             Grid Computing................................................................................................. 129

8    Pricing for Service Brokers ..................................................................................... 132
     8.1 The Impact of the Existence of Services Brokers and Maximum ..........................
          Connectivity on Price and Pricing..................................................................... 132
     8.2 A Model of an Intra-Corporate Service Exchange using CORBA ................... 134
          8.2.1    Background: Agonic Workflow Management ................................... 134
          8.2.2    Prerequisites ....................................................................................... 135
          8.2.3    CORBA Services Used to Realize the Trading System..................... 137
          8.2.4    Support of Pricing by CORBA........................................................... 138
          8.2.5    Functioning of the Proposed Model ................................................... 138
     8.3 A Model of an E-Government Portal Using .NET............................................ 139
          8.3.1    Web Services as an Enabler for E-Government Solutions................. 139
          8.3.2    Pricing of E-Government Services..................................................... 140
          8.3.3    An Example E-Government Service Broker...................................... 141
          8.3.4    Pricing E-Government Web Services ................................................ 143
     8.4 A Model for the Exchange of Standardized Software Components Using Jini 143
          8.4.1    Pricing Software Components in a Jini-Based Architecture.............. 145
     8.5 Summary: Support of Pricing Mechanisms by CORBA, .NET, and JINI ........ 148

9    A Generic Grid Architecture Supporting Multiple Pricing Models.................... 149
     9.1 Overview of the GRACE Architecture ............................................................. 149
     9.2 Implementing Pricing Models Using the GRACE Architecture....................... 152
     9.3 Adaptation of the GRACE Architecture by Grid Computing Environments.... 153
         9.3.1    TeraGrid ............................................................................................. 154
         9.3.2    NorduGrid .......................................................................................... 156
     9.4 Mapping ISP and ASP Service Models to the GRACE Architecture ............... 158
     9.5 Conclusion: Realizability of a Global Economic Market for Grid Resources.. 160

10   Conclusion ................................................................................................................. 163
     10.1 Summary ........................................................................................................... 163
     10.2 Key Findings ..................................................................................................... 164
     10.3 Future Research................................................................................................. 166

11   Appendix ................................................................................................................... 171
     11.1 Task Description / Aufgabenstellung................................................................ 171

12   References ................................................................................................................. 173

List of Figures

Figure 1: Basic Web Services Model (according to [Tsalgatidou 02]) .................................... 10
Figure 2: Types of ASP Services [Sharma 02].......................................................................... 18
Figure 3: Supply and Demand Graph ...................................................................................... 49
Figure 4: Factors Affecting Price from a Supplier’s Point of View ......................................... 63
Figure 5: Brokers Provide Consumers with Customized Services ........................................... 85
Figure 6: Components of Internet Pricing [Stiller 98] ........................................................... 109
Figure 7: Optimal Pricing Policies for ASPs [Cheng 02] ...................................................... 113
Figure 8: Internet-Based Four-Tier Architecture for ASPs [Furht 00] .................................. 114
Figure 9: Architecture of an ASP Application [Furht 00] ...................................................... 115
Figure 10: High-Level Layered Grid Architecture [Foster 03a] ............................................ 121
Figure 11: Trading Architecture Based on CORBA............................................................... 139
Figure 12: Exemplary Architecture of an E-Government Service Broker ............................. 142
Figure 13: Basic Concept of Adhesives (adapted from [Mundt 02])...................................... 144
Figure 14: Functioning of the Glueserver (adapted from [Mundt 02])................................. 145
Figure 15: An Abstract Grid Architecture [Buyya 02a]......................................................... 150
Figure 16: Draft of a Corporate IT Service Broker ............................................................... 169

List of Tables

Table 1: Categories of IT Security Services (adapted from [NIST 03]) .................................. 36
Table 2: Traditional IT Outsourcing vs. Utility Computing..................................................... 45
Table 3: Origins of Perceived Risk: Service vs. Product [Groth 95] ....................................... 52
Table 4: Characteristics of Common Pricing Approaches ...................................................... 64
Table 5: Pricing’s Requirements on IT-Infrastructure and Administration ............................ 94
Table 6: Cost Elements of the Services Provided by an ISP .................................................. 108
Table 7: TeraGrid Service Layers (adapted from [Catlett 02]) ............................................. 155

1    Introduction

The world of information technology is a domain that has a strong reputation of being
dominated by permanent change. Yet, while it is certainly true that the IT field represents a
highly dynamic environment that sees many trends and alleged silver bullets come and go,
there are, periodically, certain developments and movements that tend to have a lasting
impact. One recent phenomenon of the latter kind is the paradigm of service-oriented

Service orientation is a maxim that can currently be found throughout all fields related to
computing. However, there are two basic realms that can be distinguished. On the one hand,
service orientation is a concept that is concerned with the organizational aspects of
information technology. From this point of view, IT is regarded as a customized service that
is delivered to consumers by service providers, aiming at sparing consumers the complexity
that lies beneath their desired solutions. This is motivated by the fact that consumers are
generally not interested in the inner workings of the provided services, but primarily care
about the solution to their problems. Service-oriented IT therefore presents its products in a
manner that is understandable also for non-technicians. Becoming service-oriented must go
hand in hand with becoming more customer-oriented, ultimately leading to an increased
customer satisfaction. Furthermore, service orientation not only facilitates the communication
between IT providers and consumers, it also allows the design of more flexible solutions that
can be provided in a customized, on-demand manner. By providing well-defined services,
service providers enable their customers to have clear anticipations about what they can
expect in exchange for their IT expenses. When applied to intra-corporate IT organizations,
the concepts of service orientation turn IT departments into internal service providers that
support business, create new business opportunities, and contribute to the corporation’s
prosperity by establishing a cooperative relationship with the business functions.

The other realm in which the paradigm of service orientation has recently been evolving is
more technical. In an IT world dominated by maximum connectivity that allows the
deployment of large-scale distributed systems, software is provided in the form of services.
Similar to the organizational concepts described above, the complexity of the actual

implementation remains hidden from the user, providing him only with an interface stating
the required input and the offered output of the service. Such services can also be dynamically
composed if extended functionality is required. This concept is supported by a technology
commonly referred to as web services, which is based on a set of standards such as XML and
SOAP. Moreover, several service broker architectures (CORBA, JINI, and .NET) that manage
the matching of user requirements with available services have been developed by the
industry. Yet, the paradigm of service orientation goes deeper. With the advent of grid
computing environments, all types of resources (computational resources, storage resources,
networks, and alike) are treated as services, i.e. they are regarded as network enabled entities
that provide some capability [Foster 03]. The resources are globally available in the exact
quantity required and – again – can be accessed without knowledge of their specific
implementation. Hence, service orientation enables the virtualization of resources and thus
increases system transparency for the end-user, which in turn facilitates the user’s interaction
with complex systems. Moreover, in a service-oriented IT world, IT infrastructure can be
assembled and managed dynamically, providing resources in an on-demand fashion.

The progress in research regarding the development of service-oriented technologies has been
tremendous during the last couple of years and a common understanding of the general
concept of service orientation has well been established among the involved circles. However,
the commercial application of the emanating technologies and concepts lags behind the
technological breakthroughs. To a certain extent, this can be ascribed to a lack of coordination
between the two fields of research. Whereas technical solutions for electronic service markets
have been developed and presented, these solutions for the most part do neither take into
account economic principles, nor do they propose real-world organizational implementations.
To illustrate our point, let us formulate a number of questions that are generally not answered
by technologists, but the answers to which are crucial to commercial realization.

If services are globally available, on which markets will they be traded? Who are the
participants of those markets? What services do they offer? How are prices established? How
will the market participants influence prices? If services are created dynamically, what are the
respective pricing models? If services are customized to the requirements of consumers, will
prices be customized as well? If services are virtualized, how must real-world organizations
set their prices? Regarding the organizational aspects of the service orientation paradigm,

what are the influences on corporate IT? How can these services be managed? How are
internally provided services priced? Should IT departments aim at generating profits? How do
price and quality relate?

In the questions listed above, the pricing of services has been given a prominent role as we
believe that finding effective pricing models and mechanisms is an important step towards the
realization of service-oriented IT solutions and environments that function according to
economic principles and are therefore capable of successful commercial application.
However, service pricing must not be regarded solely from an economic or a technological
perspective, but its complexity must rather be respected by following a comprehensive
approach. Only then can questions such as the ones listed above be answered coherently.

In this thesis, we shall therefore investigate service pricing by taking into account
technological, economic, organizational and managerial issues. It is our aim to give an
overview of existing pricing models and mechanisms and put them into the context of the
available technologies. Moreover, we also want to discuss the consequences of service
orientation for organizations, putting a special focus on pricing-related issues. To do so, we
must first profoundly investigate the paradigm of service-oriented IT from the various
perspectives mentioned previously. We must then take a look at the technologies and services
that are to be priced, and present the providers who offer these services. We will then revisit
the basics of service pricing in general and the specialties of pricing IT-based services in
particular. This is followed by a discussion of the players in the markets for electronic
services and their influence on price. Ultimately, we will attempt to give an overview of the
theoretical and practical pricing models for service providers and grid computing, followed by
several drafts for architectures that can host these pricing models.

This thesis is aimed at multiple audiences. It shall serve individuals with an IT background to
gain a comprehensive overview of the possibilities and the related issues of pricing IT-based
services. Moreover, this work shall help customers of service providers to better understand
the pricing models and mechanisms they are confronted with. Furthermore, corporations that
are struggling with the quality and cost of their IT shall be able to gain an understanding of
the advantages and challenges of service-oriented IT. Finally, the academic community is
provided with ideas for further research in the pricing domain.

The further structure of this thesis is as follows:

Chapter 2 establishes a common terminology to be used in this thesis and presents a variety
of service providers and their offered services.

Chapter 3 discusses the paradigm of service-oriented IT from various perspectives. We
investigate several technologies that are relevant to service-oriented computing, followed by a
comprehensive discussion of various aspects concerning the organizational and managerial
realization of service orientation.

Chapter 4 is dedicated to the discussion of service pricing. We first review the basic
principles of supply, demand, and price, followed by an overview of common pricing policies
and strategies.

Chapter 5 discusses the specialties of pricing for IT-based services by presenting
phenomenons such as customer lock-in and network effects. Among others, we also discuss
pricing for SLA-based service provision and dynamic virtual enterprises, as well as the impact
of standards on service markets.

Chapter 6 investigates service pricing in practice. By illustrating the various market
participants’ perspectives on pricing we shall obtain a comprehensive understanding of the
market environment, so that we can formulate the requirements on pricing models and discuss
their effects on the market. This is followed by guidelines for choosing the right pricing

Chapter 7 discusses the theoretical and practical pricing models available for ISPs, ASPs,
and grid computing. Moreover, we analyze the differences between pricing various types of
resources in grid computing and illustrate the dissimilarities of pricing in scientific and
commercial grid environments. To conclude the chapter we discuss the differences between
pricing for the various types of providers.

Chapter 8 investigates the possibilities of pricing for service brokers. In order to do so we
present three generic scenarios for service brokers, each making use of a different technology,
namely CORBA, JINI and .NET. We then analyze to which extent these architectures support
pricing mechanisms.

Chapter 9 presents a grid architecture that supports the pricing of resources according to
economic principles. We investigate whether this architecture can be adapted to existing grid
environments and attempt to map the service models of ISPs and ASPs to the presented

Chapter 10 concludes this thesis by giving a summary of the work, reviewing the key
findings and outlining ideas for further research.

2     Terminology

The term service is used throughout the IT domain today and appears in many different
contexts. From web services, IT services to grid computing – services are mentioned in almost
any IT-related discussion. However, the associated meanings vary strongly and one can find it
difficult to keep up with the various concepts and models.

In the second chapter of this work, we shall therefore investigate the various types of IT-
related services that are currently discussed in IT practice and research, attempting to
differentiate more precisely among them. After having established a common terminology for
this work, we shall examine the services and the respective providers that are relevant to this

2.1       Service
According to Merriam-Webster’s Dictionary [MWOL] the term service refers to “useful labor
that does not produce a tangible commodity”. Further characteristics of services are
immediacy, high customer involvement and difficulty of standardization [Wehrli01]. Hence
traditional services cannot be stored or produced in advance because production and
consumption fall together. To express how crucial the involvement of the consumer is to the
successful delivery of a service the term prosumer (producer and consumer) has been
invented [Normann87]. The difficulty of standardization arises from the consumer’s wish for
a personalized solution, which would often inflict unbearable cost on the producer. However,
the trade-off between personalization and low cost production can be solved by offering
services that consist of standardized components. Finally, it is important to realize that almost
any product is a combination of physical goods and services. The fact that one finds physical
goods in most services and most physical goods do contain service elements complicates a
definite separation [Wehrli01]. Whether characteristics listed above are also valid for IT
services shall be examined later on in this thesis.

2.2    IT – Information Technology
The Alliance for Telecommunications Industry Solutions defines information technology as
“The branch of technology devoted to (a) the study and application of data and the processing
thereof; i.e., the automatic acquisition, storage, manipulation (including transformation),
management, movement, control, display, switching, interchange, transmission or reception
of data, and (b) the development and use of the hardware, software, firmware, and procedures
associated with this processing.” [ATIS].

In this thesis we shall use the term information technology (IT) according to this definition.
Please note that it also includes technologies used to transport information, i.e.
communication technologies.

2.3    Services in the IT Domain
From services performed by computer operating systems to the concept of e-service
provisioning, the term service is used throughout different areas of computer science and the
IT industry, covering a very wide spectrum of respective meanings. In many cases, service is
used very specifically by IT manufacturers. For example, Microsoft defines a service as “A
program, routine, or process that performs a specific system function to support other
programs, particularly at a low (close to the hardware) level.” [MSTechnet]. Whilst such a
definition - which in this case applies specifically to Microsoft operating systems (the UNIX
equivalent would be a daemon application) - may serve as illustrative examples, more general
areas in which the term service appears frequently shall be explored. In order to gain a
common understanding of the services discussed in this thesis, the following sections shall
give an overview of the various types and forms of services found in the in the IT domain.

2.3.1 E-Services
As of yet, there is no single true definition of e-services. From the product marketing
perspective, an e-service can be any electronically enabled aspect of customer utility, both in
the form of a digital purchase or an electronically mediated service after the sale in support of
previous purchases. Technologists, on the other hand, view e-services as web-delivered
software functionality, often characterized under the rubric of “web services” [Stafford 03].

                                                                                                7     E-Services from the Marketing Perspective
First, we shall briefly discuss the marketing perspective. According to [Rust 03] the
fundamental philosophy of e-service is the focus on customers and meeting their needs
precisely. Technology is considered as an enabler in e-service, but not as an end itself. [Rust
03] further defines e-service as “the provision of service over electronic networks” where e-
services can be provided by typical services organizations as well as by manufacturers of
goods attempting to strengthen their customer relationship. Major concepts enabled by e-
services are Customer Relationship Management (CRM), one-to-one marketing and e-
procurement. By providing the data and tools required to estimate a customer’s switching cost
and his potential value to the firm over the lifetime, e-services can help corporations build up
customer equity. Personalization and customization of products can be achieved by gathering
information from customers. Moreover, e-services can help corporations pursue self-service
strategies, i.e. performing a forward-integration of the customer and strengthening his role as

While e-services are easier to implement for information goods that can be distributed
electronically (e.g. software or music), e-services can also be applied in markets for physical
goods. For example, grocery stores are adding value for the customers by introducing loyalty
cards and electronic purchase tracking. The value-add to the customer consists of focused
information provision, reduced search time and increased convenience.

To sum up the marketing perspective of e-services, information-based service products and
service delivery through personalization and customization technologies build customer
relationships through superior value and higher switching-costs leading to higher customer
equity. They also provide effective means to better understand the customer needs [Rust 03].     Web Services
As stated before, from a technological point of view e-service refers to a way of delivering
software functionality across a network. This technology is commonly known as web
services. In an attempt to give a general definition for web services, we can define web
services as “self-contained, modular applications, accessible via the web, that provide a set of

functionalities to businesses or individuals” [Tsalgatidou 02]. Web services have two levels
of meaning – one conceptual (business-oriented) and one (technically) specific. Conceptually,
web services represent a model in which small pieces of application functionality are
available as services to be consumed and combined with other applications over a network.
Specifically, web services are a stack of standards that define protocols and create a loosely
coupled framework for programmatic communication among disparate systems [ATKearney].

According to [Tsalgatidou 02] a valid web services architecture requires at least the following
three roles:

   • Service provider: “A service provider is the party that provides software applications for
      specific needs as services. Service providers publish, unpublish and update their
      services so that they are available on the Internet. From a business perspective, this is
      the owner of the service. From an architectural perspective, this is the platform that
      holds the implementation of the service.”

   • Service requester: “A requester is the party that has a need that can be fulfilled by a
      service available on the Internet. From a business perspective, this is the business that
      requires certain functions to be fulfilled. From an architectural perspective, this is the
      application that is looking for and invoking a service. A requester could be a human
      user accessing the service through a desktop or a wireless browser; it could be an
      application program; or it could be another web service. A requester finds the required
      services via a service broker and binds to services via the service provider.”

   • Service broker: “This party provides a searchable repository of service descriptions
      where service providers publish their services and service requesters find services and
      obtain binding information for these services. It is like telephone yellow pages.”

Figure 1 shows a basic service model for web services. Apart from the 3 roles listed above, it
also depicts the basic activities that are needed, namely describe, publish/unpublish/update,
discover and invoke/bind.

                                             Provider        Invoke
                           Update                            Bind

                       Service                                 Service
                       Broker                 Discover        Requestor
             Figure 1: Basic Web Services Model (according to [Tsalgatidou 02])

The W3C Web Services Architecture Working Group has elaborated a set of standards to
support web services. W3C defines a web service as “a software system designed to support
interoperable machine-to-machine interaction over a network. It has an interface described in
a machine-processable format (specifically WSDL). Other systems interact with the Web
service in a manner prescribed by its description using SOAP-messages, typically conveyed
using HTTP with an XML serialization in conjunction with other Web-related standards.”

The three main standards underlying the W3C web services architecture are:

   • SOAP (Simple Object Access Protocol) is a protocol for calling and executing services
      on a remote system. It is based on XML and uses the HTTP protocol to carry its data.

   • WSDL (Web Services Description Language) is a XML-based description of how to
      connect to a particular web service.

   • UDDI - (Universal Description, Discovery and Integration) represents a set of protocols
      and a public directory for the registration a real-time lookup of web services.

[Ferris 03] describes the basic functional principle of web services:

“In a service-oriented architecture [...] the service provider has a service designed for others
to use. The provider creates a WSDL service description that details the interface, that is, the
operations of the service and the input and output messages for each operation. A binding
implementation description for the services is then created that describes how to send each
message on the wire where the service is located. The WSDL now contains all the information
needed to invoke the service. The service provider now publishes the WSDL service
description to one or more discovery agencies. Typically the role of the discovery agency will

be fulfilled by a registry, such as UDDI, that allows additional information describing the
hosting business and makes associations with the taxonomy to be published along with the
WSDL description so that others can find the service using a wide variety of search criteria,
including category-based searches. Eventually, the service requester finds the service
description via the discovery agency. It then uses the WSDL description to develop or
configure a client that will interact with the service through the service provider.”

One of the key applications of web services is Enterprise Application Integration (EAI). Web
services can “glue together” applications running on two different messaging product
platforms, enable database information from one application to be made available to others
and enable internal applications to be made available over the Internet [Kreger 03]. Web
services are also an enabler for service-oriented architecture, which shall be introduced in a
later section of this work.    Traditional Services vs. E-Services
Finally, we shall analyze whether the characteristics defined for traditional services apply also
to e-services. Traditional services are considered intangible due to the fact they are
performances rather than objects [Hoffmann 03]. It is therefore difficult for consumers to
evaluate the value of a service until they have actually experienced it. According to
[Hoffmann 03] e-services are able to overcome intangibility issues to a certain extent by using
web-based channels to provide evidence of service such as appearance of a website, frequency
of information updates and automated order and shipping confirmations.

Traditional services are marked by immediacy, meaning the simultaneity of service
production and consumption. Moreover, the producer and the consumer must both be
physically present in order for the transaction to be completed. While the basic issue of
immediacy is valid also for e-services, consumer involvement in the actual production process
can be minimized by leading consumers step-by-step to control the flow of the process, which
standardizes the process and increases the efficiency of the transaction [Hoffmann 03].
Because physical distance between producer and consumer is hardly an issue for e-services,
and many e-services can be accessed from mobile devices, constraints regarding the
geographical position of the producer and the consumer do not apply for e-services.

While describing the characteristics of traditional services, it was pointed out that services
cannot be produced in advance or stored in warehouses, hence they are perishable. Whilst this
also holds true for e-services, the perishability issue is somewhat eased by the fact that most
e-services are available 24/7 due to their highly automated production. Moreover,
mechanisms to flatten out high variations in supply and demand can also be more easily
implemented for e-services than for traditional services.

2.3.2 Intra-Corporate IT Services
A business oriented IT department sees the facilities it offers its users as services. An IT
service can then be defined as a set of related functions provided by IT systems in support of
one or more business areas. This service can be made up of software, hardware,
communication facilities as well as manual labor, but the users perceive it as being a self
contained, coherent entity [ITSMF].

The scope of services required from and performed by an IT department varies strongly from
organization to organization and the establishment of a comprehensive listing of possible
services would exceed the scope of this thesis. However, [Peppard 03] provides a possible
categorization of the services that an IT department may provide in order to support the
business. The categories are as follows:

   • Applications services refer to those services delivered via software applications. These
      services are derived from the information handling abilities of technology and include
      information processing services, information sharing services, information storage
      services and information access services. Application services directly impact the
      performance of business processes with process designs or components of processes
      imbedded in software applications.

   • Operational services are those services that relate to assembling and operating the core
      IT environment. Such services include installation of hardware and software,
      maintaining   the   communications      network       and   servers,   upgrading   software,
      configuration management, change management, trouble-shooting hardware and
      software problems, and running the data centre.

   • Value-enabling services are services that are provided to enhance the value of
      information assets or identify opportunities provided by IT to better manage
      information. Examples include IS strategy development, systems analysis, systems
      design, requirements gathering, infrastructure architecture, network design, user support
      (including helpdesk), purchasing, vendor management, and consulting.

   • Infrastructure services are those services that are derived directly from the
      infrastructure investment, essentially the technology itself. In fact, these services are
      better described as technical capabilities rather than services in the true sense of the
      word. These IT capabilities are provided by the hardware, software and communications
      infrastructure of the company. On their own, they provide little direct value to the
      company — without them, however, application will have nothing to run on. These
      services include capacity (bandwidth and storage), connectivity, scalability and
      flexibility. Security can also be included in this category.

Depending on the size of a corporation, the sheer amount of necessary services can yield
enormous demands on IT management. Not only must services be provided, but the IT
department must ensure that they are reliable, efficient and correct. Providing this type of
service requires a dedicated and focused infrastructure consisting of the hardware, software,
communication, documentation and skills required to support the provision of quality IT
services. The management of this infrastructure to provide the required services at the
required level of quality is called service management [ITSMF].

2.3.3 The IT Services Industry
Not all IT services required by a corporation’s business functions may be provided internally,
but are rather purchased from external service providers. These providers make up the IT
services industry. A complete listing of all available services on the market is not the aim of
this thesis. Instead, we shall take a closer look at the services offered by the types of providers
that are of central importance to this thesis, namely Internet Service Providers (ISPs),
Application Service Providers (ASPs) and providers of grid computing services.

                                                                                                13    Services Provided by Internet Service Providers
An Internet Service Provider (ISP) is a company that provides individuals and companies
access to the Internet and other related services. ISPs fall into three broad groups: backbone
providers, national providers and local providers [HP 99].

  • Backbone providers are nationwide or multinational organizations that control Internet
     routing. They often own significant pieces of the backbone itself.

  • National providers buy capacity and routing services from backbone providers and
     operate Points of Presence (POPs, local access points to the Internet) across the country
     (or the world). National providers are often described as resellers since they are simply
     reselling bandwidth that they have purchased from the backbone provider.

  • Local providers operate in the same way as the national group, but on a smaller scale.
     Usually they work within a smaller geographic area.

According to [Rao 00] the services offered by ISPs can be split into 3 categories:
connectivity, basic services and value-added services.

  • Connectivity services allow the user to connect to the ISP’s infrastructure in order
     access the Internet. Consumers will normally be provided with dial-up access to their
     ISP using their standard telephone line, though this is starting to change as cable
     organizations (and others) provide higher capacity lines using other access methods.
     Commercial and larger organizations may use other connection methods which offer
     greater bandwidth and provide “always on”' services. The most common access is by
     leased lines offering a permanent connection to the Internet.

  • Basic services offered by ISPs include access to the World Wide Web (WWW), Usenet
     (Newsgroups), and electronic mail (e-mail). The following are generic, high-level
     descriptions of these services that shall support the creation of a common understanding
     of the respective terms.

  • Internet access: The ISP operates the necessary infrastructure to provide authorized
     clients with access to Internet services such as the World Wide Web (www), Telnet or
     File Transfer Protocol (ftp).

  • Electronic Mail: The ISP maintains the necessary infrastructure and manages user
     accounts so that authorized clients can send and receive e-mail messages by using the
     respective software on their side.

  • Usenet: The ISP maintains the necessary infrastructure and manages user accounts so
     that authorized clients can send and receive newsgroup messages by using the
     respective software on their side.

Apart from these basic services, ISPs have introduced numerous other services, so-called
value-added services, of which the most important shall be described in short.

  • Hosting: The business of housing, serving, and maintaining files for one or more
     websites is called (web-) hosting. ISPs may offer subscribers free space for a small
     website that is hosted by one of their servers. Virtual hosting implies that the service is
     transparent and that each website carries its own domain name and set of e-mail
     addresses. Dedicated hosting is the provision of a dedicated server machine which is
     dedicated to the traffic to the customer’s website. Companies may purchase their own
     servers and place them at a site that provides fast access to the Internet and is operated
     by an ISP. This practice is called colocation.

  • Virtual Private Network (VPN): A virtual private network is a way to use a public
     telecommunication infrastructure, such as the Internet, to provide remote offices or
     individual users with secure access to their organization's network. The goal of a VPN is
     to provide the organization with the same capabilities that owned or leased lines offer,
     but at a much lower cost [Interoute].

  • Voice over IP (VoIP): By providing Voice over IP services, ISPs allow their customers
     to use the Internet as a transmission medium for telephone calls. This means that voice
     information is sent in digital form in discrete packets rather than in the traditional
     circuit-committed protocols of the public switched telephone network, avoiding the tolls
     charged by the ordinary telephone service [MAC].

  • Web-based e-mail (webmail): Webmail services allow users to access their e-mail
     accounts via a web browser instead of designated mail client software. This is
     convenient for traveling users, for they have global access to their e-mail account and
     do not have to adjust their specific server settings when using a public terminal.

Practical Example: ISP1
“PensionCo” is one of the most recognized brands in the life, pensions and investments
industry and employs approximately 50,000 staff. There are 2,500 “PensionCo” branches
servicing approximately 16 million customers worldwide.

When “PensionCo” decided to offer its services via the Internet to consumers and partners, it
turned to a renowned ISP that could provide a stable, secure and scalable hosting platform to
host the entire “PensionCo” e-commerce infrastructure. Security and system stability were
seen as key factors along with the requirement for an infrastructure that could grow with the
clients’ and the industry’s demanding requirements.

The selected ISP was able to provide “PensionCo” with a high availability solution, which
included complete system redundancy and managed operations to ensure the e-commerce
network would experience the minimum possible downtime, an issue seen as the highest
priority by “PensionCo”. The hosting solution incorporated complete 24/7 management and
provision of the infrastructure within a secure and remote location offering further levels of
physical security. A scalable architecture that would support the projected growth of
“PensionCo’s” e-commerce network was also delivered.

The ISP also provided a VPN solution that allowed “PensionCo” field agents to securely
connect to the corporate network while residing at customer sites. Moreover, an extranet
portal for “PensionCo” business partners was introduced to allow the facilitated and secure
exchange of confidential documents.

The implementation phase of the project was completed within one year, with a five year
contract between “PensionCo” and the ISP for the further management of the provided
infrastructure. “PensionCo” rates the solution put in place as a great success, as it has allowed
“PensionCo” to expand its business through the company’s web site, intranet and extranet.

 The following shall serve as an illustrative example of a successful ISP deployment. The author of this work is
fully aware that this is an idealized example and not all ISP related projects can be completed in such a
successful manner.

                                                                                                              16    Services Provided by Application Service Providers
An Application Service Provider (ASP) is a third-party entity that manages and distributes
software-based services and solutions to customers across a wide area network from a central
data center [Webopedia].

ASPs provide business solutions to companies on a rental basis, i.e. enterprises rent
applications instead of buying [Lande 00; Wittmann 00]. Rather than spending millions of
dollars on a new enterprise-wide software implementation, the organizations pay the ASP on
a per-transaction or per-month basis for hosting, running and managing the software. ASPs
have complete responsibility for managing the infrastructure, hiring manpower, and helping to
conduct business online [Sharma 02].

Many ASPs have specialized on providing solutions for a certain branch of industry, whereas
others offer services that can be applied in a multitude of industries. [Dewire 00] and [Focacci
03] define five major categories of ASPs, based on the types of services they offer:

   • Enterprise ASPs usually offer high-end applications that require customization. This
      category of ASPs offers application solutions such as ERP, SCM, CRM or workflow
      software services. They are both horizontal (expertise in functional applications in many
      industries) and vertical (expertise in specific industry applications and solutions) in
      terms of customer target, as their primary focus is the management of enterprise-level

   • General Business ASPs focus mainly on the needs of the small to mid-sized companies.
      Thus, they primarily deal with general business applications that require little or no
      customization. Because of this, most have a horizontal customer focus.

   • Specialist ASPs focus on a particular type of application, such as human resources
      management software, and provide the software application, management and related
      support and services required to completely manage the business functional area(s) they
      choose to serve. Specialist ASPs tend to have a horizontal approach. The difference
      between this model and the general ASP business model is that specialist ASPs provide
      more depth and breadth in terms of addressing the aspects of the business function(s) in
      which they specialize.

  • Vertical ASPs implement software applications and manage related solutions and
     services that are targeted for specific industries. These ASP’s primary focus is on the
     customers in their targeted industry.

  • ASP aggregators (AASPs) bring together numerous underlying ASPs and are designed
     to provide customers with a wider range of software applications suites and related
     services than a single ASP alone.

The most common applications provided by ASPs include Customer Relationship
Management (CRM), Enterprise Resource Planning (ERP) and Messaging and Collaboration.
CRM and ERP applications are often very complex solutions so that organizations that lack
the required skills and resources to manage this kind of software may find it more convenient
to outsource to an ASP. Moreover, CRM applications often need to be accessed by mobile
users (e.g. sales representatives), which is another reason especially for smaller companies to
use the services of an ASP that can provide the necessary infrastructure to support dial-in
users. System availability and reliability is the main reason for organizations to outsource
their messaging and collaboration applications to ASPs that may possess dedicated hardware
and networking infrastructure to guarantee high availability. ASPs can also provide
businesses with temporary online cooperation solutions or file-sharing possibilities that may
be required for project work but are not worth a long-time investment.

            Collaborative Services           Electronic Commerce                 Content Services

             shared/virtual workspaces,         auctions, catalogues,         published text, TV and radio
             chat/threaded discussion,         transaction processing,           programming, games,
            videoconferencing services,      government services, direct       events, music, community
            shared calendar/scheduling,       marketing, EDI, logistics,      created content, educational
              document management,          personnel/corporate financial         courseware hosting
                  training services                services, survey

                                Corporate Systems / Knowledge Management

              ERP, sales force automation, project management, knowledge management, archive indexing,
                  policy-based information access management, claims-processing, e-forms/workflow

                   Interfaces                  Networked Smart                     Infrastructure
                                                  Products                          Outsourcing
            portals, search engines, in-      devices that use network              server/software
                box management,             connections to send out their       administration, web site
             personalized information        own calls for service when         hosting, help desk, call
                       access                 required, wireless locator     centers, directory/addressing/
                                                      services                messaging administration,
                                                                                 remote management

                             Figure 2: Types of ASP Services [Sharma 02]

According to [Sharma 02] the services offered by ASPs can be classified as collaboration
services, electronic commerce, content services, corporate systems / knowledge management,
interfaces, networked smart products and infrastructure outsourcing. Figure 2 depicts this
categorization and lists the respective applications.

Practical Example: ASP2

“TELCO” an IP telephone provider that had undergone explosive growth in its four-year
history had found itself in need of ERP application in replacement of its out-dated PC-based
accounting system. “TELCO” had surpassed the maximum number of users allowable on its
software licenses, and the business/financial employees had to share their core applications.

Although current needs could have been handled by less complex software, “TELCO”
decided to implement a solution based on SAP R/3 or PeopleSoft because further expansion of
the company was predictable and “TELCO” didn’t wish to spend time and money migrating
the data off an interim solution later on.

But “TELCO” could not afford the amount of more than $500,000 it would cost to buy and
install its own PeopleSoft system. Additionally, “TELCO” was lacking the required human
resources required for the project. “TELCO” therefore decided to turn to an ASP from which
it could rent the PeopleSoft software on a monthly basis, instead of buying and installing it by
its own means.

“TELCO” paid a one-time fee for the installation, which included on-site training and moving
its existing data to the new system. Additionally, “TELCO” pays the ASP a monthly usage-
based fee per user. “TELCO’s” 30 PeopleSoft users access the application via a virtual private
network on a dedicated T-1 line. According to “TELCO”, no network upgrades were needed
to accommodate the new application.

                                                                                       (Continued on next page)

  The following shall serve as an illustrative example of a successful ASP deployment. The author of this work is
fully aware that this is an idealized example and not all ASP related projects can be completed in such a
successful manner. The example was adapted from [Paul 00].

Because “TELCO” was a rather young company, it didn’t have any major legacy systems or
legacy business processes. “TELCO” therefore did not have to customize its PeopleSoft
application, with one exception: The ASP’s team added the ability to capture serial numbers
of components within its telephony gateways in order to provide proper servicing of those

“TELCO” rates the project as a success, mainly because of the smooth transition and the fact
that it does not have to devote its staff to the complex ERP system. Instead, the ASP’s experts
manage “TELCO’s” PeopleSoft solution, while “TELCO’s” staff can focus on its core
business.     Services Enabled by Grid Computing
A grid computing system is a distributed parallel collection of computers that enables the
sharing, selection and aggregation of resources. This sharing is based on the resources’
availability, capability, performance, cost and ability to meet quality-of-service requirements
[Franklin 03]. From a more business related point of view, grid computing can be described
as the transparent sharing of IT resources that may reside at geographically distributed sites.
For a good introduction to grid computing please refer to [Berstis 02].

When we look at the services enabled through grid computing, services cannot be defined as
accurately as in the case of ISPs and ASPs. The reason for this is that grid computing is
currently still in the developmental phase and commercial providers of grid computing
services are only emerging. A vision for grid computing is to see grids as the enabler for
turning computing services into a utility. Computing power would then be accessed “from the
plug”, similar to electricity. [Wladawsky 02] illustrates this analogy by asking whether it
makes sense for companies to all manage their own complex IT infrastructure, and reasons on
whether it would be possible to outsource these assets and activities to specialized “power
plants” (data centers), as in the case of electric power generation. As a possible realization of
such a vision is still years away, numerous grids operating in closed environments have been
implemented. However, standards for grid computing, especially for the interconnection of

such (e.g. the Open Grid Services Architecture (OGSA) [OGSA]), are still in development
and most of the grids already deployed implement proprietary standards [Tanner 03]. The lack
of a common standard is one issue that is currently preventing a large-scale deployment of
commercial grids. Other major concerns for most potential grid users are security, the lack of
applicable pricing and accounting mechanisms, as well as the fact that most applications must
be redesigned in order to be grid-compatible.

[Berstis 02] discusses several capabilities of the grid computing architecture, the most
important of which are listed in the following:

   • Parallel CPU capacity: By splitting up computation-intensive problems into “subjobs”
      that are distributed among the nodes of a grid and executed in a parallel manner, one
      can obtain massive computing power that may even exceed the performance of
      monolithic supercomputers.

   • Optimized resource utilization: In today’s IT environments, many resources are only
      utilized during a fraction of the time. Grid computing can help utilize idle machines and
      optimize the overall computational load; when a machine experiences an unusually high
      load, jobs may be routed and run on an idle machine elsewhere on the grid. However,
      [Berstis 02] lists two prerequisites for this to be possible. First, the application must be
      executable remotely and without excessive overhead. Second, the remote machine must
      be capable of running the application, i.e. it must meet any special hardware, software
      or other resource requirements imposed by the application.

   • Virtual collaboration: Grid computing can support virtual collaboration by giving
      individuals access to distributed resources. A “data grid” can span multiple systems
      such as databases and provide users with increased reach. By implementing intelligent
      data allocation mechanisms, data transfer rates can be optimized. Sharing is not limited
      to files, but may also include other resources such as equipment, software, services and
      others. These resources are then “virtualized” in order to provide a more uniform
      interoperability among heterogeneous grid participants [Berstis 02].

   • Reliability: In conventional architectures reliability is mostly based on expensive
      hardware redundancy. In grid computing there is still hardware redundancy, however
      this redundancy is provided by the other nodes of the grid, i.e. there is no need for
      designated hardware redundancy components. Therefore reliability relies more on
      software reliability. If a system fails, the respective jobs can be redirected to another
      node of the grid. For critical applications, multiple copies of important jobs may be run
      at different locations.

These capabilities of the grid computing architecture may be combined to provide innovative
and possibly commercial services. As of today, most grids are used for academic research
purposes. Some commercial grids exist, however these are mostly limited for use within the
owning organization. As stated above, the vision for grid computing goes much further. The
idea is that individuals and organizations will offer their unused CPU cycles and other IT
resources to other participants on the grid. Services would be virtualized and offered via the
grid and in an on-demand fashion. A company experiencing increasing demand for resources
might first respond with spill-over services from its own IT infrastructure, and if that’s not
sufficient, it may access additional capacity from other grid participants, purchasing only as
much capacity as is needed at the moment [Wladawsky 02]. For such a vision to become
reality, major challenges have to be overcome. If grids are to interconnect, the individual
participants have to be sure that the confidential data they send over the grid cannot be
accessed by third parties. This may also be a psychological hurdle that has to be mastered and
it is certainly doubtable whether corporations will ever be comfortable with sending
confidential intellectual property over open networks. The other major issue to be resolved is
the lack of applicable pricing, accounting and billing models and mechanisms for grid
services. As long as the grid participants cannot properly account for the use or provision of
resources, they are lacking a strong incentive to participate in an open grid. The existing
pricing and market models for grid computing are mostly theoretical and their usability in
practice has yet to be proven. Pricing mechanisms for grid computing shall be discussed in
detail later in this work.

Practical Example: Grid Computing3

“Biocorp”, a biotech firm, uses computer software to simulate the behavior of the human
body and predict its response to various drugs. A pharmaceutical company approached
“Biocorp” with an idea for a new drug, and by way of a special application “Biocorp’s” team
of programmers and scientists simulates the drug's effect on hundreds of "patients" who take
the medication in hundreds of different circumstances. Each trial can involve up to 13,000

Software this complex can overwhelm even the fastest computer. According to the company's
chief software architect, running that many simulations on a single server would take two
years. “Biocorp” therefore decided to build a computing grid which allows “Biocorp” to
borrow processing power from 145 different machines spread across its offices and run all
13,000 simulations in less than a week.

The “Biocorp” grid consists of 125 servers and 20 desktops connected by an ordinary local
area network. The grid is operated by two different management applications: one which
handles the servers and a separate one for the desktops.

A central management console, loaded anywhere on the network, oversees the grid, constantly
trading messages with small software agents running on 145 client machines. Given a set of
tasks that need completing, the console determines which clients have available processing
power and distributes the tasks accordingly. When a task is complete, the client sends the
results back to the console. The advantage of this arrangement is that even while the grid is in
operation, employees can still use the client machines for their regular work.

Setting up the grid was not an easy task. The purchased grid software had to be customized by
“Biocorp” in order to achieve compatibility with “Biocorp’s” database that holds all of
“Biocorp’s” data for the drug simulations. However, even the employees with little computer
experience have no trouble using the grid. The company's programmers also customized its
central console, adding a particularly simple and straightforward interface.
                                                                          (Continued on next page)

    adapted from [Metz 03]

“Biocorps” drug tests are a nearly perfect application for a computing grid. Divided into
thousands of self-contained simulations, they are easily spread across multiple machines,
reducing the overall computing time dramatically.

2.4    Summary
As we have demonstrated in the preceding sections the usage of the term service in the IT
domain comprehends many different aspects. If one is to create a high-level categorization,
one can distinguish between two domains of services. On the one hand, there are services that
are performed in order to support the operation of an IT environment, whereas on the other
hand services may also be provided with the help of IT infrastructure. While the former
contains services that are provided with the help of manual labor, the latter are mostly purely
electronic services. Services that are performed in order to support an IT environment shall be
in the following be referred to as IT services. IT services will be discussed in the following
chapter where the service-concept is inspected from managerial, communicative and
economic perspectives.

3     The Paradigm of Service-Oriented Computing from
      Various Perspectives

Service-oriented computing, also known as service-oriented IT, is currently one of the most
discussed topics in the field of information management. Within the increased service
orientation, we can identify two major developments. On the one hand, IT organizations
within corporations are regarding themselves the longer the more as internal service
providers, treating the business functions as customers. On the other hand, service orientation
is also a new paradigm for the development and deployment of software and other
computational resources. Emerging from component based software engineering, web
services are currently the state-of-the art concerning the development of distributed

These two developments have pervasive consequences for the entire IT domain. In order to
describe the effects and changes as comprehensively as possible, we will in the following
discuss the new paradigm of service-oriented computing from several perspectives.

3.1    Technological Perspective on the Paradigm of Service-
       Oriented Computing

3.1.1 Service-Oriented Architecture (SOA)
Many of today’s corporations are forced to manage a very heterogeneous system
environment. While the reason for the heterogeneity may lie in the historic growth of the
corporation, organizations may also find themselves forced by their specific requirements to
follow a best-of-breed approach (i.e. selecting new applications in regard to the most suitable
functionality but not necessarily to the best compatibility), hence further raising the
complexity of their IT environment. As the interoperability of applications becomes crucial
for many businesses, the integration of various systems can represent a serious challenge. The
following section presents the principles of service-oriented architecture (SOA) which is
based on the communication between software components and potentially enables
corporations to integrate IT resources.

In a service-oriented architecture the enterprise business systems and applications are
considered services which are available on a network. IT infrastructure is abstracted so that
the functionality is presented in the form of loosely coupled services which can be accessed
by users independent of the underlying technology that supports them. Many existing IT
architectures today are n-tier, meaning that the infrastructure is distributed, separating data
business logic and presentation logic onto different systems. SOAs introduce an abstraction
layer to such n-tier infrastructure. Two forms of abstraction are used in an SOA:
encapsulation and virtualization [Zapthink 03]. A software object is said to be encapsulated
when its inner operations are hidden from the outside (this is also known as the “black box”-
principle). Virtualization on the other hand, describes the procedure of having multiple
heterogeneous resources appear as one homogenous entity to the user.

SOAs implement the concept of component-oriented software. In component-oriented
software, applications are assembled from a set of pre-existing components. A software
component is defined as “a piece of software with one or more well-defined interfaces that are
configurable, integrable, and not modifiable” [Hofmann 99]. In this scenario, applications are
built by combining a set of components that communicate with one another by means of
shared interfaces. This requires an appropriate component infrastructure that provides
mechanisms for searching the needed components, obtaining component information,
component communication and component assembly. SOAs have the same requirements;
however the granularity of SOAs is coarser than that of software components, meaning that
SOAs encapsulate software components (that may have been built using different
technologies) or whole applications and virtualize them into services. A service has a
network-addressable interface, stresses interoperability and may be dynamically discovered
and used [Stevens 02].

Today, several technologies that are capable of realizing a service-oriented architecture exist.
CORBA (Common Object Request Broker Architecture), Microsoft DCOM (Distributed
Component Object Model) and web services (introduced in a previous chapter) all provide the
necessary functionality. However, [Zapthink 03] argues that regarding the realization of a
SOA, web services are superior to CORBA and DCOM technologies due to the web services’
loose coupling and use of open standards. Loose coupling depicts the separation of the
participants in distributed computing interactions so that modifying the interface of one

participant does not influence the other. This is especially important to the abstraction in the
form of encapsulation described above. The combination of loose coupling and the existence
of commonly accepted standards allow the creation of services without any knowledge of the
consumers of those services [Bloomberg 03].

One of the advantages of SOAs is the reduction of the complexity of integration of
heterogeneous environments by encapsulating existing functionality and exposing it as a
service [Zapthink 03]. This not only facilitates the enterprise-wide integration of applications,
but also enhances data exchange between business partners. Corporations may also achieve a
better return on investment on their software by being able to re-use certain components for
different applications. Further advantages of SOAs are support for multiple client types, code
mobility (due to location transparency and lookup services the location of the service is not of
importance to the client), better scalability and higher availability (a load-balancer may
forward client requests to multiple service instances) [Stevens 02].

3.1.2 Mobile Agents and Peer-to-Peer Computing: Alternatives to the
       Client-Server Model?    Mobile Agents
The idea of mobile agents is not new and has been discussed long before the phenomenal rise
in popularity of the Internet. However the recent developments in the web services technology
have again drawn attention to this concept. [Chess 94] describes mobile agents as “computer
programs [...] dispatched from a client computer and transported to a remote server computer
for execution”. Mobile agents can migrate from one system to another during their execution
and communicate amongst one another, clone, merge and coordinate their computations.
Mobile agents are autonomous agents in the sense that they control their relocation behavior
in pursuit of the goals with which they are tasked [Meng]. Main fields of application for
mobile agents are information retrieval on the www, distributed database access, parallel
processing, automation of electronic marketplaces and others.

Mobile agents can be implemented using the previously described web services standards. By
using SOAP messages mobile agents can interact with other agents and take advantage of

more specialized content they hold. WSDL can be used by mobile agents to describe their
capabilities, and UDDI can provide directory services to speed up the discovery of other
agents [Fou 01]. In addition, [Huhns 02] lists several points by which mobile agents extend
web services:

   • A web service knows only about itself, but not about its users, clients or customers.
      Agents are often self-aware at a meta-level, and through learning and model building
      gain awareness of other agents and their capabilities as interactions among agents occur.

   • Web services, unlike agents, are not designed to use and reconcile ontologies.

   • Agents are inherently communicative, whereas web services are passive until invoked.

   • A web service, unlike an agent, is not autonomous.

   • Agents are cooperative and by forming teams and coalitions can provide more
      comprehensive services. Current standards for web services do not provide composing

[Lange 99] explains that in the traditional client-server model, the client needs some
information about the server to decide which services it can use. Clients may make use of
Remote Procedure Call (RPC) to evoke a procedure on a remote server. RPC is synchronous;
meaning that after sending a request to the server the client is suspended until it receives the
results of the call. While the procedure is being performed on the server there is on-going
interaction and communication between the server and the client, thus inflicting traffic load
on the network.

Contrary to RPC, mobile agents do not call a procedure on a remote server, but supply the
procedure itself. After being dispatched, the mobile agent operates autonomously and
asynchronously without having to maintain a permanent connection to the client. This not
only saves bandwidth, but is also suitable for scenarios where connectivity is of poor quality
and / or expensive, e.g. in the case of mobile devices.

A further advantage of mobile agents is their dynamic adaptability; they can sense their
execution environment and react independently to changes. Multiple mobile agents have the
unique ability of distributing themselves among the hosts in the network to maintain the

optimal configuration for solving a particular job. Mobile agents’ ability to react dynamically
to unfavorable situations and events makes it easier to build robust and fault-tolerant
distributed systems. If a host is being shut down, all agents executing on that host are warned
and given time to dispatch and continue their operation on another host in the network unlike
RPC where the transaction might remain unfinished [Lange 99].

A major concern for mobile agents is security. Because the agents execute code on the target
server, servers must be protected from malicious agents and vice versa. Mechanisms to verify
the integrity of the executable code as well as the authenticity of the agent and the server are
therefore required. Another way of increasing security is by deploying mobile agents only in a
trusted environment, such as an intranet or a virtual private network (VPN).

One can argue that most of the applications for mobile agents can be implemented using
traditional client-server architecture. However, the aggregate technical advantages of mobile
agents as well as the rise of the web services standards make this technology the longer the
more attractive for the use in distributed applications.    Peer-to-Peer Computing
A technology that shares some similarity with mobile agents is called peer-to-peer computing
(P2P). In P2P networks, every participant node acts both as a client and as a server. Data is
distributed among the peers instead of being concentrated on a central server. The similarity
of P2P and mobile agents lies in the way search and navigation are handled. In P2P, the peers
propagate the requests similar to the way that network nodes propagate mobile agents.

[Parameswaran 01] describes the basic difference between P2P networks and client-server

“A client-server scenario like the web depends on a single server storing information and
distributing it to clients in response to their requests. The information repository remains
essentially static, centralized at the server, and subject only to updates by the provider. Users
assume a passive role in that they receive, but do not contribute, information. A P2P network,
on the other hand, considers all nodes equal in their capacity for sharing information with

other network members. Each user makes an information repository available for
distribution, which, combined with anyone’s ability to join the network, leads to the growth
stemmed primarily from the fast growth of a network composed of distributed information

According to [Milojicic 02], P2P has more decentralized control and data compared to client-
server systems and provides better support for ad-hoc connectivity and anonymity. One must
be careful when comparing performance and fault-resilience of client-server and P2P systems.
Individually, client-server systems have highly optimized performance and fault-tolerance
which cannot be claimed for the individual clients taking part in a P2P network. However, if
one looks at the aggregate systems, P2P has higher aggregate performance and fault resilience
by avoiding dependency from individual servers and having fewer points of failure. Because
P2P is based on shared ownership, the cost of ownership is distributed among the peers and
therefore lowered for each individual.

Applications for P2P networking are parallel processing, collaboration and information / file
sharing. The latter is certainly the most popular form of P2P computing, however due to its
widespread misuse for illegal trading of copyrighted material, it is also the most controversial.
This leads us to the disadvantages of P2P networks. The main challenges when operating a
P2P environment is the release of control. From the technical point of view, this is especially
critical regarding security and scalability. In an open system, where anyone can participate,
maintaining trust and content integrity becomes crucial and improving the overall system
performance can be very challenging when there is no central server to optimize or upgrade.
From a non-technical point of view the release of control is critical regarding the monitoring
of the shared content. Content is added and removed dynamically, as peers connect and
disconnect from the network. To keep unwanted content off the network self-regulating
communities may be introduced, however the provided anonymity in many P2P networks
makes it difficult to track down and penalize misbehaving users.

The success of a P2P system depends on network effects, i.e. the phenomenon that a service
becomes more valuable as it is being used by more people4. The network effects themselves

    For an extensive discussion of network effects please refer to [Shapiro 98]

depend on the acceptance of use. Different from a client-server system, a P2P network ceases
to exist when it is abandoned by its users.

Unfortunately, not all users provide a benefit for their peers and some users can actually
damage the system. Freeriders, i.e. users that download content without providing any
content, do not create any network effects for their peers, and individuals with poor resources,
e.g. low bandwidth, can slow the network down by becoming bottlenecks for the system as a
whole [Milojicic 02]. In economics, this phenomenon is known as “tragedy of the commons”
and describes a situation where the careless use of a free, shared resource makes the resource
unusable for all, in this case erasing all network effects.

The question arises whether any quality of service can be guaranteed in a P2P system. The
dynamic nature of P2P based on the unreliable and untrusted components that make up the
system make the establishment of QOS guarantees very challenging. [Kubiatowicz 03]
describes a series of methods that can guarantee certain properties for P2P systems, however
the guarantees provided are of probabilistic nature and not as strong as the ones that can be
given for centralized or client-server systems.    Conclusion
Both peer-to-peer (P2P) as well as mobile agent technologies have the potential to substitute
client-server solutions in many areas. Mobile agents provide powerful ways of processing
large amounts of data by bringing the application to the data, instead of transporting the data
to the application. Their partial autonomy makes mobile agents distinct for use in mobile
devices. Peer-to-peer has proven to be superior to the traditional client-server architecture in
at least some aspects, illustrated by the great success of file-sharing networks. However, the
client-server architecture is well established and its strengths lie in the well-understood
programming models, guaranteed performance, strong security and well-known solutions for
reliability [Milojicic 02]. While there may be many exciting new applications to be realized
with the help of P2P and mobile agents, there are just as many domains where client-server
architectures can and will provide solutions that are just as convenient or even superior. For
many systems connectivity is rather secondary while security is crucial. It is therefore very
unlikely that the client-server model will vanish in favor of P2P or mobile agent technologies,

but will rather remain a key architecture for many distributed computing environments.
Client-server, P2P and mobile agent systems have all empirically proven their right to exist
and the future will most likely hold a coexistence of all three architectures.

3.2    Managerial Perspective on the Paradigm of Service-Oriented
Over the past decades IT departments within corporations have emerged from their existence
as supporting functions within accounting departments to becoming key business functions
that are crucial factors for the success of many corporations. In modern organizations there
are hardly any activities remaining that are not somehow influenced or supported by IT. With
its growing importance to the business, IT has become a major budget item, with many IT
managers finding relatively easy approval for IT projects and enlarged IT budgets during the
years of Internet- and IT hype at the end of the 20th century. However, with the hype being
over, IT spending has become a political issue in many corporations and IT departments are
often considered a central source for cost savings. As for any other department, managers
responsible for IT must be able to show the value their department provides to the
corporation. Moreover, IT processes and services must be liable to the same quality
requirements as all other business units. On the other hand, it is necessary that the services the
IT department provides are charged to the respective users.

It emanates from the previous section that the activities and outputs of the IT department are
regarded as services. As discussed in a precedent section of this work, corporate-internal IT
services most often involve manual labor and are performed to support the different business
functions by providing well-functioning information systems, with the ultimate end being the
increase of a corporate efficiency and effectiveness.

3.2.1 Quality Management: Service Management Frameworks and SLAs
A key aspect of the management of IT services is ensuring high quality of the services
provided. Quality is usually based on the conformance of expectations. There are two
dimensions that determine whether or not an actual service meets the user expectations
[Peppard 03]:

      • Technical quality: this dimension is concerned with what the user receives and when

      • Emotional quality: this dimension relates to how the customer experiences the delivery
         of the service

In order to ensure that IT services effectively support the business, service management
frameworks have been introduced5. Such frameworks are sets of documents that define best
practices for IT services provision and management. The aim is to help IT departments
become more customer and service-oriented. Another benefit of introducing a service
management framework is the establishment of a defined, common terminology which by
itself may lead to a reduction of the misunderstandings that are often experienced in
discussions between IT staff and lines of business. Ultimately, service management
frameworks should lead to precise specifications of the services that IT performs in favor of
the business, which is a premise for quality measurement. Only if the expected services are
precisely defined it can later be measured whether those aims are achieved and the services
were performed in a satisfactory manner.

Service Level Agreements (SLAs) may be established between the IT services provider and the
business functions. The purpose of SLAs is to define the performance required of the service
and to put in place measurement mechanisms whereby actual performance against targets can
be monitored [Parish 97]. It is important that SLAs contain targets that are quantifiable and
therefore objectively measurable. SLAs are helpful to service providers as well as to the
customers. They help the customers by showing them which levels of service they may expect
and provide them with a basis for their complaints if the expected quality is not delivered.
Service provider on the other hand can use SLAs to configure their services to the specified
needs of a singular customer and are at the same time protected against ever-growing
customer demands regarding their efforts. If quality is to be ensured, it is necessary that the
    For an introduction to an example of a service management framework please refer to [ITIL]

providers regularly assess their outputs with regard to the service levels specified in the SLA.
Existing quality gaps must be identified and discussed with the customers so that the actions
that are required to ensure quality of service can be taken.

3.2.2 Cost Management: Chargeback
According to [Sun 02] chargeback is the process of breaking down the costs associated with a
single task or project and then distributing these costs back to the organizational consumers.
As stated at the beginning of this section, it is important that the services performed by IT
departments are charged to the respective users to ensure fair cost allocation [Oleson 98].
However, a precise allocation is very difficult and many times subject to political issues. In
mainframe environments, users are often charged for the MIPS (Million Instructions Per
Second) they have consumed. In practice, full costs of the IT department are often calculated
into the price of a MIPS. Another way is to charge each business unit a fixed price per
workstation (often referred to as a per seat charge), where again total IT costs are calculated
and divided by the number of users, resulting in a fixed amount that departments are charged
per user. The advantage of these procedures is their simplicity; however the drawback lies in
the lack of transparency. As it is one of the aims of a service-oriented IT environment to
provide customers with high transparency of the value they receive in exchange for their
expenditures, other charging mechanisms have been introduced. Instead of charging a fixed
price per seat or MIPS, IT departments put together a catalogue of services to which business
units may subscribe. The customer is then charged either a fixed or variable price for the
services he has consumed. Whether a fixed or a variable price is charged depends on the
nature of the service. For some services it may be possible to charge the effective usage
whereas for others this is not practicable or the overhead for measuring the actual use would
simply be too large. It is also common to offer a service at a fixed price at certain service
level, with improved service levels being available at increased prices.

The dilemma of service-oriented charging lies in the trade-off between offering customized
services that provide a high cost-transparency and the need of implementing complicated
mechanisms that are able to measure the effective usage. IT management has to decide on
how far the provided services can be customized without producing an excessive
measurement overload. Moreover, IT departments must decide on how much insight

customers should have into the services. As previously mentioned, the services to be provided
should be precisely described in SLAs, however it is questionable whether the customer
should also be confronted with the inner workings of a service. Some customers may want to
have a detailed look at the processes that make up the individual services, whereas others may
simply expect the overall system to function in the agreed manner. How much insight to
provide may be a political matter, as IT departments that offer too much insight may find
themselves having to debate on charging matters on the process-layer, whereas the ones that
present their services as “black-boxes” may be accused of not providing the required billing
transparency. It is therefore the IT management’s task to find an appropriate abstraction level
that provides enough pricing transparency but at the same time avoids having to justify single
work steps.

3.2.3 Management of IT Risk and Resilience
As we have mentioned throughout this chapter, IT plays a crucial role for nearly every
business today. With business heavily relying on properly functioning IT services, the
associated risks have also become substantial. Hence, IT risk management must be of high
priority in order to assess and minimize the potential threat to business.

[NIST 01] defines risk as a “function of the likelihood of a given threat-source’s exercising a
particular potential vulnerability, and the resulting impact of that adverse event on the
organization.” According to [NIST 01] IT risk management is the process that allows IT
managers to balance the operational and economic cost of protective measures and achieve
gains in mission capability by protecting the IT systems and data that support their
organizations’ missions. [NIST 01] divides IT risk management into three processes: risk
assessment, risk mitigation, and evaluation and assessment. Risk assessment includes
identification and evaluation of risks and risk impacts. Risk mitigation refers to prioritizing,
implementing, and maintaining the appropriate risk-reducing measures resulting from the risk
assessment process. Evaluation and assessment describes the continual process of evaluating a
successful risk management program.6

  A complete review of IT risk management would exceed the scope of this thesis; interested readers are kindly
referred to [NIST 01]

Following the service-oriented point of view of this thesis, let us now discuss the services that
are related to IT risk management. Risk management itself can be regarded as a high-level
service, consisting of the processes listed above. However, from a more practical approach,
risk management will most likely result in the installation of IT security services. Hence,
security services are provided and consumed in order to minimize the risks identified by an IT
risk assessment and are therefore part of the IT risk management realm.

[NIST 03] provides a categorization for IT security services. As depicted in Table 1, three
categories – management services, operational services, and technical services – can be
differentiated. Table 1 also gives examples for services in each category.

    Category             Category Description                   Examples

    Management           Techniques and concerns normally       IT security program
    Services             addressed by management in the         development, IT security
                         organization’s IT security program.    architecture, certification and
                         Focused on managing the IT             accreditation
                         security program and the risk within
                         the organization.

    Operational          Services focused on controls           Contingency planning, incident
    Services             implemented and executed by            handling, testing
                         people (as opposed to systems),
                         which often rely on management
                         activities and technical controls

    Technical            Services focused on security           Firewalls, intrusion detection,
    Services             controls a computer system             Public Key Infrastructure
                         executes. These services are
                         dependent on the proper function of
                         the system for effectiveness.

            Table 1: Categories of IT Security Services (adapted from [NIST 03])

With respect to our previous discussion of service providers, we must note that most IT
security services can be performed by either internal or external partners. For example, it is a
common practice to have external service providers that possess the respective know-how
perform security assessments of a corporation. Recently, offerings for the outtasking of
services that fall into the technical category have also appeared. [NIST 03] gives the example
of intrusion detection systems (IDS) where outtasking may well be reasonable, because IDSs
require a high level of expertise to operate properly and need large, frequently updated
databases to be able to detect the most recent threats. According to [NIST 03], external
service providers may have greater leverage to accomplish this at less cost because several

customers can share the overhead of maintaining the database. There may well be other
examples of IT security services in which lack of know-how or potential for economies of
scale make outsourcing to an external provider favorable. However, as for any outsourcing
decision, the associated risks must be profoundly analyzed. In the case of security services
this risk analysis becomes crucial, as the risks involved with the outsourcing of security
services can often be critical to the mere existence of a corporation. Moreover, strict service
level agreements as well as close cooperation between the outsourcer and the outsourcing
provider(s) are crucial in order to be able to react swiftly in case of an incident. The selection
of an outsourcing partner must therefore be performed with the utmost care.7

The term resilience has been widely used recently to describe a corporation’s ability to
recover from negative external influences. [Anderson 03] defines resilience as an
organization’s ability to adjust easily to change and states that the key to resilience rests in
maximizing the ability of systems and processes to effectively support a business under any
adverse, fast changing, or unexpected condition. With regard to our previous discussion,
resilience can therefore be seen as the ultimate aim of all risk management and security
related activities. From a managerial perspective, strong resilience can only be achieved if the
processes within an organization are highly interoperable both between the different lines of
business as well as the supporting functions. Hence a corporation’s IT must be installed in a
way that allows sufficient functionality even if part of its system have been interrupted or
damaged. This can be achieved by both organizational (e.g. by including redundancy) as well
as technological (e.g. supported by evolving technologies such as autonomic computing)

Returning to the paradigm of service orientation, we can specify the requirement on IT that all
activities and measures related to the achievement of high resilience must be transparent to
the user. Hence, the user should not be confronted with security- and resilience-related
decisions and issues, but should rather be provided with secure and highly available services.
Whereas the provision of such a service may not (yet) be possible on the level of individual
information systems, IT as a whole should present itself to the user as an integrated and
autonomic system that provides the required services in a highly resilient manner.

 A comprehensive guide to IT security services, including checklists to support the selection of a provider is
given by [NIST 03]

3.2.4 Management of Web Services
The preceding argumentations in this section have dealt with corporate IT services in general.
However, the focus shall now be put specifically on the management of web services, as this
technology is becoming more and more important for many corporations.

Of course, web services are subject to the same managerial criteria as any other services
provided by an IT organization, i.e. web services must also are also liable to quality measures,
their cost must be monitored and their usage must be charged. Whereas frameworks for IT
services management have been established and standardized, web services management
standards and frameworks are only evolving8. However, [Seth 03] describes the architectural
and organizational components that are required in order to manage web services effectively.
From the managerial perspective, [Seth 03] lists the following components to be of primary

     • Contract Management: “Contracts are defined for linking the web services clients to the
         terms of service (these terms include billing and rating packages). This contract is
         associated with each subscription of web services clients. When the clients invoke the
         web services, authentication, authorization, and identity systems match the associated
         contract. Web services clients will not have access to invoked Web Services in case of
         an invalid contract.”

     • Metering and Billing: “Clients will access web services based on the contract they have
         agreed upon when they subscribed to web services. To bill the clients, the system
         should be aware of the statistics of the contract, SLA, and quality of service (QoS)
         promised to the clients. A metering application gathers this usage and session statistics
         and the billing application is responsible for co-relating the above-mentioned two
         statistics and produces a bill to the clients. Therefore, the contract management and
         metering application should be treated as inputs to the billing application.”

    for an exemplary web services management framework please refer to [HPWSMF]

   • Auditing: “This web service management component is built on monitoring and logging
      components. The basic functionality it offers is to produce audit trails enabling the
      reconstruction and examination of a sequence of events in web services context. The
      auditor fulfills this function by reliably and securely keeping records of security-related
      events. Various security event data can be obtained from the logging application.
      Security events could include authentication events, policy, enforcement decisions, and
      so forth. The resulting audit trails may be used to detect attacks, confirm compliances
      with policy, deter abuse, or other purposes.”

3.3       Communicative Perspective on the Paradigm of Service-
          Oriented Computing
Unfortunately, communication between IT organizations and business functions is often
characterized by misunderstandings, arguments and conflicts. Possible topics of such disputes
may be services that do not meet expectations, imprecise service charging or the assignment
of competencies. Furthermore, IT managers often have difficulty with demonstrating the
value their organization provides and discussions whether IT should be run as a cost- or a
profit center have also become apparent. These issues shall be inspected in the following and
it shall be checked whether an intensified service orientation may provide solutions in this

3.3.1 Common Areas of Dispute
The most common source for conflict lies in the lack of a shared language between business
and the IT organization. While the language of the IT representatives can be very technical,
business has its own specific vocabulary that may be tangling to outsiders as well. In order to
avoid misunderstandings that may result in unsatisfactory service delivery it is important that
a common, shared language is established between IT and business. As mentioned previously,
this can be achieved by introducing a service management framework which defines
important terms and phrases. A common vocabulary is also a premise for the establishment of
SLAs, because the setting up of targets is useless if the involved parties do not share a mutual
understanding of these targets. Once SLAs are successfully established they can strongly
contribute to the improvement of communication and therefore cooperation between business
and IT. SLAs serve as a basis for the discussions between the two parties and function as
insurance for both sides. The IT organization can protect itself against unjustified reproaches

regarding the delivery of services and the business functions have a contract to which they can
refer if the services delivered by IT are poor. The amount of discussions about the scope and
quality of the delivered services can therefore be reduced, freeing up resources on both sides
to focus on their respective core activities.

Many conflicts between business and the IT departments arise because service-oriented
thinking is not firmly rooted in the IT organization. If an IT department truly considers itself
as an internal service provider it must see the business functions as its customers and provide
them with the same treatment an external customer would receive. With IT functioning as a
cost center that does not have a revenue target but rather holds a quasi-monopoly for internal
service provision such a mindset may indeed be difficult to assert. However, the establishment
of SLAs containing penalty clauses as well as permanent service controlling and comparison
with outsourcing possibilities may apply the necessary pressure on the IT department to
achieve a more customer-oriented behavior.

When a business department is charged for certain services it will demand cost transparency.
As described earlier in this work, providing total transparency may be very difficult,
sometimes even impossible, in the case of IT services. Offering a service catalogue that
enables variable pricing is certainly more advantageous than rough-estimate per seat fixed
charges, however if complete transparency cannot be provided, it is important that the IT
organization communicates its reasons. If IT can explain that a fixed price is in the end more
favorable for the customer than having to pay for expensive measuring mechanisms that are
required for a precise cost allocation, the drawbacks in transparency will certainly be accepted
up to a certain level.

Business often underestimates the complexity of IT architectures and therefore has unrealistic
expectations regarding the flexibility of the IT organization. While it is certainly true that
business functions can expect IT to perform in the needed and specified manner, business
must also understand that service specifications and requirements cannot be altered
permanently without influencing service quality. However, if IT can design its services in a
modular fashion, flexibility may be increased, hence enabling quicker responses to new or
altered service requirements.

Overall, a service-oriented IT organization should facilitate communication between IT and
business. By giving clear definitions in a common language of the services that are provided,
IT can anticipate misunderstandings and increase customer satisfaction. It is important that
communication between IT and business is an on-going process and does not only occur when
problems arise. While business depends on IT for maintaining its competitiveness, IT’s mere
existence is based on serving the business. A partnership based on cooperation and open
communication is therefore crucial. Finally, the two parties should share the same ultimate
goal, which is the prosperity of the one organization they are both a part of.

3.3.2 Communicating the Value of IT
It is well possible that during the years of IT-frenzy many corporate IT departments may have
grown out of proportion and have lost focus on their original purpose which is to support the
business. However, one must never forget that an IT department’s right to exist is based on its
function to help its corporation increase its effectiveness and efficiency. Maintaining an IT
department simply for the sake of IT cannot be justified. As IT is the longer the more
regarded as a commodity by individuals that underestimate the importance of proper IT
management, Chief Information Officers (CIOs) may find themselves in the difficult situation
of having to justify the cost and value that their department provides to the corporation. The
justification may be difficult because IT investments tend to have rather medium- to long-
term returns on investment, meaning that an immediate pay-off is often not apparent for many
IT projects. This is due to the fact that IT usually provides tools which increase the
effectiveness or efficiency of business processes and such improvements are often difficult to
measure and verify. Providing corporate decision boards with the hard numbers they often
demand may therefore be nearly impossible. [Pender 00] recommends that CIO’s demonstrate
the value of their projects by showing how they align with overall corporate strategy and
suggests describing the situation in which the company will find itself if the investment in the
project is not made. [Brennan 02] argues that the value of an IT project is best described from
the three different perspectives of the customers, the employees and the shareholders. While
in the eyes of the customers a technology initiative may be worthwhile if it makes a company
easier and more convenient to deal with, the most worthwhile initiatives to the employees are
those that make their jobs more interesting or give them more time to perform more value-
added tasks. Shareholders on the other hand may give most attention to the return on
investment, i.e. questioning whether the project will provide economic value by improving

quality, cutting costs or increasing revenue. [Brennan 02] recommends that before an IT
investment decision is made all three perspectives should be evaluated, giving most weight to
the shareholders’ argumentation.

3.3.3 The IT Department: Cost or Profit Center?
The running of an IT department causes costs. These costs must be allocated to the business
units as described in the previous section. Traditionally it is not an IT department’s aim to
generate profit, but rather to allocate all its costs. It therefore functions as a cost center.
However, this is no ultimate classification and it is worth a thought whether an IT
organization could not also be run as a profit center. The difference between a cost- and a
profit center is that a profit center contributes directly to a company’s revenue. It must
therefore have access to a market where it can sell its products and services. Traditionally, IT
doesn’t directly contribute to a corporation’s profit – as long as productivity gains or mission-
critical functionality are not factored in [Bernard 03]. But with the technological advances
described in previous sections of this work this may no longer have to be the case. One could
argue that because IT provides a company with new communication channels to reach its
customers, the revenue generated through these channels (e.g. revenue generated by online
sales) should be assigned to the IT organization. Such an argument may be used to underline
the undisputable importance of IT to the business, but it remains rather weak and can be
replied with a counter-example of facility management: hardly anyone would argue that the
revenue generated in a certain building should be assigned to the facility management
department because it provides the company with the means to do business. IT shall be seen
as an enabler, and apart from companies that operate in the IT industry, IT is a supporting
function of the business and therefore not obliged to generate profits. However, if again the
vision of “utility computing” presented earlier in this work is picked up, this may indeed
change in the future. [Hapgood 03] argues that companies whose core business may not lie in
providing IT related services could offer excess computing or storage capacity on a respective
market and generate revenue. However - as stressed earlier - such a vision, if not to say
utopia, lies yet in the far future and for now IT organizations shall be considered cost centers,
without underestimating the overwhelming value they provide to modern corporations.

3.4     Economic Perspective on the Paradigm of Service-Oriented
When the idea of IT outsourcing was first introduced many years ago it usually concerned
hardware and facility infrastructure. As the focus shifted over the years, software also became
an outsourcing topic; beginning with contract programming during the 1970’s and reaching to
the rise of the previously described Application Service Providers (ASPs) in the late 1990’s9.
ASPs provide software as a service delivered over a network such as the Internet.
Implementation, integration and management of the software can therefore be outsourced to
the ASP. From an economic point of view, ASPs achieve cost efficiency by maintaining
goods and services for specialized organizations and maximizing learning curve effects and
economies of scale. The customer’s benefits of being partnered with an ASP lie in the ability
to acquire valuable resources that are important to production and distribution, without
substantial investment [Lee 03]. The investment curve is therefore flattened, meaning that
instead of a large, one-time investment in software followed by increasing cost for software
maintenance, the payment streams are leveled out in the form of periodic service payments to
the ASP. This makes software cost more predictable, however the dependencies on the ASP
should not be underestimated.

Web services were introduced earlier in this work. The ASP model and web services have in
common that they both provide software in the form of a service. However, while ASPs take a
monolithic software application, place it on a remote server and host the software as a service
over a network, the web services model is more distributed and based on the composition of
applications out of a number of software components that may be running on different
computers. Yet, to take full advantage of the potential of the web services model, mechanisms
for dynamic service composition are required. Without going into technical detail, such an
architecture requires a service directory (e.g. UDDI), a service description language (e.g.
WSDL) and a standardized format for message exchange (e.g. SOAP)10.

While service discovery and binding may be realized in a closed environment using the
described web services standards, the idea of having a market for commoditized electronic

  For an overview of the IT outsourcing evolution please refer to [Lee 03]
   For the description of dynamic service composition using Jini, Java-Beans and XML technologies please refer
to [Mennie 00]

services yet remains a vision for the future. In such a market, service requestors would be able
to lookup the required service in a directory, choose among a number of adequate service
providers and request the service from the provider that best suits the current requirements. It
is also possible that service brokers would provide service requestors with the most
convenient service available by choosing from a range of service providers according to the
criteria submitted by the service requestor. Such brokers could exist if the transaction cost
were substantial. Economically speaking, intermediaries (i.e. brokers) would be efficient, if
the cost inflicted on the service requestor and the service provider where higher than the sum
of the transaction cost inflicted on the broker [Benjamin 02]. However, for such a market to
be realized, several obstacles have yet to be mastered. The fundamental question that has to be
answered is whether web services can truly become and be treated as commodities. In order to
be considered a commodity, a good must be homogenous, i.e. it is standardized to the extent
that there are no differentiating features remaining. With the standardization of web services
in progress, one can assume that eventually platform portability issues will be resolved and
web services that offer standard functionalities may indeed become commodities. Technical
issues aside, [Benjamin 02] lists the following three areas in which advances need to be made
if a true, open market for web services is to become reality:

      • Payment mechanisms that enable subscribers (publishers) to pay for (derive revenue
          from) web services on a per-invocation basis. These mechanisms must also be
          convenient for the processing of very small amounts, i.e. so-called micro payments.

      • Billing mechanisms that are capable of tracking the usage of web services on a per-
          invocation basis and provide variety of tariffs corresponding to different performance
          ratings and service levels.

      • Anonymity can be an obstacle when transactions are to be made between parties that are
          previously unknown to one another. However, anonymity may also be required to
          enable re-branding and protection of reputation. This could be achieved by involving a
          trusted third party in the transaction.

Further issues that need to be resolved include the establishment of trust among the market
participants11, security and fraud protection as well as system reliability. These issues are not
limited to technological matters and therefore must be discussed from organizational and

     Please refer to [Bolton 02] for an analysis of the role of trust in electronic markets

social point of views as well. So for now we can conclude that while the required theoretical
models and technical standards for the realization of a commodity market for electronic
services are available, several technical and non-technical issues remain to be resolved, if
such markets are to become reality.

We have mentioned the idea of utility computing several times in this work so far and we
shall now inspect this concept from an economic perspective. In utility computing, not only
software but also hardware resources are offered in the form of services. According to the
idea of utility computing, organizations no longer will purchase hardware equipment, but
rather subscribe to services that provide them with the required computing, networking or
storage resources. Utility computing is therefore also a form of IT outsourcing. However, if
the utility computing concept is compared to traditional IT hardware outsourcing there are
some substantial differences. First of all, while traditional outsourcing is usually based on a
long-term contract between an organization and a single outsourcing provider, the
relationships among service consumers and providers are more dynamic in utility computing.
Ideally, contracts are negotiated for short-term, “on-the-fly” service delivery so that service
consumers do not bind themselves to a single provider, but rather select the provider that is
most convenient at the time the service is needed. The risk of lock-in is therefore reduced.
Another difference lies in the pricing schemes. Traditional outsourcing contracts usually
require fixed periodical payments. As in the case of traditional utilities such as water or
electricity, organizations implementing utility computing are charged on a pay-per-use basis.
This leads to a conversion of fixed costs into variable costs.

Table 2 sums up the basic differences between traditional IT hardware outsourcing and the
utility computing model.

    Traditional IT Outsourcing               Utility Computing

    one provider                             pool of providers

    Long-term contracts                      Ad-hoc, short-term contracts

    Fixed cost, periodical payments          Variable cost, pay-per-use

    Risk of lock-in high                     Risk of lock-in reduced

                   Table 2: Traditional IT Outsourcing vs. Utility Computing

3.5    Conclusion
In the preceding sections, we have discussed a number of issues regarding the organization of
IT services within a corporation. Reviewing the discussion, we can identify three main groups
of stakeholders: corporate-internal IT, business functions, and corporate management.

We have argued that a key issue of today’s IT organizations is the increased cost pressure
they face, combined with an augmented need for justification of their cost-value ratio. Hence,
IT organizations are in need of tools and methods that allow clear communication of the
services they provide to the business, and the cost at which the provision of these is possible.
For the purpose of becoming a service-oriented organization, IT departments should therefore
make increased use of tools such as SLAs and service management frameworks, which allow
the establishment of a well-defined service catalogue. Corporate-internal IT then presents
itself as a service provider, offering an IT service portfolio from which business units can
select the required services that optimally support their operations.

Business functions are in need of IT services in order to operate efficiently and stay
competitive. In general, business expects IT to provide the required services in the respective
quality, without having to bother about technical details of the provided solutions. However,
many IT departments are unable to provide this service, often resulting in conflicts about
service quality and costs between the two parties. A further issue for business is that it is often
not clear what service or value is actually received in exchange for IT expenditures, i.e. cost
transparency is not given. To improve this situation, the establishment of clear service
definitions and prices in a mutual language between business and IT is required. Such
installations not only increase the efficiency of communication between business and IT, but
also allow better control of IT-related costs.

We assume that the main interest of corporate management is economic prosperity of the
organization. With regard to our discussion, efficient cooperation between business and
supporting functions such as IT is therefore of key importance. It is the responsibility of the
management to install an organizational setup that allows the optimization of the
corporation’s IT service portfolio. Hence, internal IT must be organized in a way that the
services required by the business can be provided (either internally or acquired from an

external source) in the needed quality at the lowest possible cost. The management of the IT
portfolio then needs to consider – amongst others – the technical, managerial, and economic
advantages and issues of service orientation we have discussed in this chapter. We will
present an idea of how such an organization could be realized by adapting the service broker
concept in section 10.3, in which proposals for future research are discussed.

As we have seen in this chapter, corporations are in need of mechanisms that allow proper
charging and accounting for all IT services in order to implement their IT in a service-oriented
fashion. This requires the introduction of feasible pricing models that are both accepted by
users as well as providers. Yet, the correct introduction of such a model requires the
understanding of pricing, a domain which is characterized by its enormous complexity. In the
following chapters we shall therefore provide comprehensive insight into the field of service

4     An Overview of Service Pricing

The fourth chapter of this work is devoted to the various aspects of pricing. Before we look at
the specialties of pricing IT-based services in the next chapter we shall first review the basic
economic principles of price, followed by an investigation of the different factors that affect
price in a market. By reviewing common pricing policies and strategies we then present
different approaches to service pricing. To conclude the chapter we present a framework to
grasp the different aspects of pricing.

4.1    The Basic Principles of Pricing
According to [Rowley 97] price has three functions in a market from a macro-economic
perspective, namely allocation or rationing (the balancing of quantities demanded and those
supplied); stimulation and acting as an incentive for new players and products to enter a
marketplace; and distribution, whereby income is distributed between buyers and sellers.

From a micro-economic point of view, i.e. the view of the individual organization, price
represents the value at which a seller is prepared to trade, and the value at which the customer
agrees to participate in that exchange, respectively. In economic models, individuals are
assumed to act in accordance with their ultimate aim of maximizing their individual profit.
From the seller’s position, profit can be expressed as:

       Profit = (Price x Quantities sold) – Total cost

While this equation obviously omits the complex and dynamic mechanisms of pricing
strategy, consumer behavior etc. it demonstrates the close relationship between profit and
price. According to the equation, profit is higher the higher price can be raised, assuming
stable unit cost and sales volume [Rowley 97].

4.1.1 Supply, Demand and Price
In classical economics, the concept of supply and demand is used to determine a so-called
equilibrium price. Supply refers to the varying amounts of a good that producers will supply
at different prices; in general, a higher price yields a greater supply. Demand refers to the
quantity of a good that is demanded by consumers at any given price. According to the law of

demand, demand decreases as the price rises [Klein 83]. Price can then be regarded as the
balance between supply and demand. The concept of supply and demand can be illustrated
with a graph, as depicted in Figure 3.



                        Equilibrium Quantity

                               Figure 3: Supply and Demand Graph

The described model applies to pure commodity markets with completely undifferentiated
products, which is a premise that can hardly be found in reality. [Rowley 97] lists further
assumptions that underlie the model, but do not comply with real markets. The model:

   • assumes that the buyer has near perfect information about the alternatives and the
      market; in reality a buyer is rarely that well informed.

   • assumes that competing brands are reasonably close substitutes for one another. In real
      markets, suppliers devote considerable effort to strategies such as branding to
      differentiate their products, or encourage consumers to view their product as different –
      in some way – from those of their competitors.

   • assumes that the demand curve is known; this is hardly ever the case because suppliers
      are rarely in a position to be able to test the market at different price levels.

   • takes no account of the ongoing costs of a purchase, such as, for example, subsequent
      subscriptions or maintenance costs, and focuses only on the initial cost.

According to the model of supply and demand, consumers and suppliers act as price takers,
i.e. they cannot actively influence price but can only decide if they want to buy or sell for the
given price. In reality, suppliers develop a pricing strategy which is a key element of every
marketing mix. The process of setting a specific price for a good or service is called pricing.

4.1.2 Factors Affecting Price
There are numerous internal and external factors that have to be considered when pricing
decisions are made. [Rowley 97] categorizes these factors into organizational factors,
customer factors and market factors.

Organizational factors influencing price decision-making are related to the organization’s
objectives and resources, with regard to the existing product portfolio and the organization’s
potential to influence that portfolio. When making pricing decisions organizations have to
consider their products’ life-cycles and preferably create a well-balanced product-portfolio. It
is also possible that matured products are used to finance a penetration pricing strategy
(discussed later in this work) for new products.

Very important internal – and therefore organizational – factors are costs. Obviously, costs
have a direct impact on profit, as we have shown in the precedent section. The total costs
inflicted on an organization by the production of a good or a service are composed of variable
and fixed costs. Variable costs vary directly with the level or production, whereas fixed costs
are independent of the quantity produced. As the aim of most any organization is to at least
recover cost, price has to be set appropriately. In order to make a profit, a business should
ensure that its products are priced above their total average cost. In the short-term, it may be
acceptable to price below total cost if this price exceeds the marginal cost of production so
that the sale still produces a positive contribution to fixed costs [Tutor2u].

Customer factors influence price because of the impact their aggregated demand has on price.
For scarce goods, an increase in demand will lead to increased price. However, this factor’s
relevancy is limited in markets where there is a virtually unlimited supply, e.g. in information
markets. Yet, customers always have an impact on price as they will determine whether a
price is acceptable or not by comparing the benefits they will receive from the trade with the

cost inflicted upon them, thus limiting an organization’s ability to raise prices. A premium
price (compared to competition) may be set if there is a perceived additional customer benefit
or strong brand loyalty, usually based on quality. Some pricing models that will later be
presented take this into account. With regard to information markets, [Boulding 68] argues
that the value of information is not readily quantifiable and impossible to determine in
advance. Accordingly, the concept of customer benefit cannot easily be applied to information
markets. Nevertheless, [Rowley 97] states that buyers will always have an acceptable range of
prices for a specific product type, such as the time that they are willing to invest to locate
specific information or the amount they are prepared to pay in order to have this time reduced.

Important criteria for consumers when choosing a product or service are brands. Brand names
are usually associated with a certain quality or service level, having an impact on the
consumers’ price expectations. Depending on the image of their owned brands, producers of
goods and services can choose different pricing strategies. Owners of premium brands, i.e.
brands that have the reputation of above-average quality, may charge superior prices, whereas
others may attempt to create a brand that stands for low, discount prices.

Market factors that influence price include competition and the market environment.
Competition is a major factor because price makers need to take into account the prices set by
competition. Not only should the producers of the same product be regarded as competitors
but also producers of similar products and substitution goods. The market environment
consists of social, technological, economic and political factors that shape the marketplace in
which a producer operates and therefore influence price. Specific examples for such factors
are government-regulated prices or the general economic situation such as a general recession
where public and consumer spending underlies strict constraints usually requires price cuts.

The pricing of services differs from the pricing of products to a certain extent. Earlier in this
work (in section 2.1), we have presented the general properties of services. With regard to
pricing, it is important to realize that the perceived risk associated with the purchase of a
service is generally higher for services than for physical goods. A higher risk leads to a
reduced perceived value, thus lowering the consumers’ willingness to pay and therefore
limiting the service provider’s possible profit margin. The main reason for the higher

perceived risk is that compared to physical goods services are more difficult to examine and
compare before the time of purchase. Table 3 lists further reasons for a higher perceived risk.

                Product                         Service

                Observable                      Output not observable

                Attributes observable and can   Attributes not observable
                be demonstrated
                Can be tried, operated, felt    Uncertainty of attribute-needs
                Utility is in the product       Uncertainty about provider’s
                Hidden intent can exist         Uncertainty about provider’s
                Dire comparisons can be         Not necessarily the best value
                made, product can be tried
                out before purchase
                Returnable                      Non-returnable

             Table 3: Origins of Perceived Risk: Service vs. Product [Groth 95]

In order to reduce the perceived risk and therefore be able to set higher prices, service
providers may offer guarantees. Moreover, SLAs (as we have discussed in section 3.2.1) may
be used to give precise descriptions of the offered services to the customers, and therefore
reducing possible gaps concerning the quality of the service between the provider and the

In summary, pricing is a matter of matching internal and external factors with regard to both
marketing as well as financial considerations. Because these factors are numerous and may be
contradicting, pricing strategies which define consistent pricing policies have to be defined.

4.1.3 Pricing Policies
There are numerous possible pricing policies that may be selected to effectively support a
chosen pricing strategy. The following short overview of common pricing policies is to a
large part adapted from [Rowley 97].

Pioneer Pricing Policies

Pioneer pricing policies are concerned with setting the base price for a new product.
Organizations need to consider development costs and the time period over which they aim to
recoup these costs, and how easily and quickly competitors can enter the market. There are
two basic approaches, although in practice, many producers will choose a compromise
between price skimming and penetration price:

   • Price skimming is charging the highest possible price that buyers who most desire the
      product will pay. The advantage of price skimming is that development costs can be
      covered relatively quickly, and demand can be kept consistent with production capacity.

   • Penetration price is a price set below the prices of competing brands in order to
      penetrate a market and produce a larger unit sales volume. The objective of penetration
      pricing is to gain a large market share quickly. The disadvantage is that subsequently
      the organization has less flexibility with pricing because, while it is easy to lower
      prices, the market is normally very intolerant of raised prices (unless the initial price is
      seen as a promotional price).

Organizations sometimes use penetration pricing after skimming pricing so that they achieve
a large market share if competitors are expected to enter the market quickly.

Price Discrimination

Price discrimination is the act of charging different prices for the same product or service to
different groups of consumers for reasons not associated with costs. Price discrimination
serves the skimming of the consumer surplus. The discrimination can be applied according to
different criteria [Wehrli 01]:

  • Horizontal price discrimination: the overall market is split up into separate market
     segments which each contain consumers that possess similar willingness to pay. The
     resulting different prices for the same product or service may lead to obvious
     disadvantages for single customers, depending on market transparency. This type of
     price discrimination must therefore be combined with further product or service
     differentiations (brand, product, technology, way of distribution, etc.).

  • Temporal price discrimination (e.g. different tariffs during day and night) serves
     primarily capacity balancing.

  • Spatial / geographic price discrimination can be applied when regional markets differ
     from one another. This type of price discrimination is most likely to be applied for
     consumer goods.

  • Personal price discrimination according to buying behavior may be applied to bind
     profitable customers to an organization’s products and services.

Psychological Pricing

Psychological pricing encourages purchases based on emotional rather than rational
responses. There are a number of well-established approaches, which have varying validity in
different marketplaces:

  • Odd even pricing – odd pricing, for example, $99.95 to avoid psychological barriers;
     even pricing $32.00 rather than $31.95 in order to communicate extra quality.

  • Customary pricing, or pricing at a price that the customer has always paid, such as 10
     cents for a phone call. With customary pricing a producer may alter the service offered
     for the price rather than the price, as when the length of telephone access available for
     10 cents changes.

  • Prestige pricing – for example, where prices are set high to create a prestige image, e.g.
     for luxurious consumer goods.

Professional Pricing

Professional pricing may not be related directly to value or what the customer is prepared to
pay, but may be determined by custom and practice within the industry. For example,
prescription charges are set at a standard rate, conventions for estate agents are that the fee is
a percentage of the selling price of the house; fees for dentists are determined by an
established scale. Professional pricing is most evident in the IT industry for consultants and
others offering services, such as information intermediaries. Here, professional pricing may
be on the basis of the job done, according to an agreed contract or on time taken. Others may
offer comparable services “free” with a view to attracting business or retaining internal
customers. This clearly affects the price that others may charge for a similar service.

Promotional Pricing

Special pricing tactics may be adopted in association with a promotion that is designed to
draw attention to a specific product. In the IT industry this is most evident with promotions on
special configurations of hardware. The options are:

   • Price leaders – are priced below usual mark-up, at a level near cost, or below cost;
      revenue from other products or services should offset these cuts.

   • Special event pricing which is intended to increase sales volume and generate operating

4.1.4 Common Approaches for Service Pricing
A vast amount of pricing models and methods has been published in economic literature. A
complete listing would certainly exceed the scope of this work. The pricing approaches that
shall be presented in the following have been chosen with regard to relevancy for this work.
The previously described pricing policies are of general nature and can be applied to any
product that is to be sold, may it be a physical good or a service. As the focus of this work lies
in service pricing, the common approaches for price determination that are presented in the
following can be applied specifically to service pricing. The overview is based on [Tung 97],
but has been extended with further pricing models.

Cost-Oriented Pricing Approach

In cost-oriented pricing the seller determines the cost caused by the provision of a specific
service and adds the desired profit-margin to calculate price. Cost-oriented pricing is usually
based on full cost, but it can also be a contribution and incremental basis. The advantage of
this approach is its simplicity, which makes it a quick and easy to use pricing method.
However, the drawbacks are that it does not consider supply and demand, does not maximize
profit and does not include unique service characteristics and selling conditions into the
pricing decision [Hoffman 89].

Competitive-Oriented Pricing Approach

In competitive-oriented pricing, price is determined with reference to the prices of the
competitors, i.e. price is set to meet the market’s competitive situation. This approach is
usually chosen when objectives are to increase sales or market share. The major advantage of
this approach is its simplicity, however it provides no guidance on how much higher or lower
than a competitor’s price a service provider should set its price [Arnold 89].

Extended Cost-Oriented Pricing Approach

[Hoffman 89] proposes an extended cost-oriented pricing approach for professional service
providers. The model includes the traditional cost-oriented pricing factors of fixed costs,
variable costs and the organization’s profit goals, along with the factors that make up the
extended model: essentiality (the extent to which the purchase of the service is postponable),
durability, value added, and the percentage of performance capacity. The factors influencing
service pricing can be partitioned into three parts:

   • traditional factors: variable cost (VC), fixed cost (FC) and profit goal (PG)

   • unique premium characteristics: essentiality (Ep), durability (DURp) and tangibility
      (value added, V*)

   • percentage of performance capacity used

To formulate a service price, first, a service manager needs to determine a proper cost basis
(FC + VC) for a particular service. Second, the cost used in the pricing basis should be

adjusted by the percentage capacity of fixed cost actually consumed for delivering a service.
For example, if a service performance consumes less than 100 percent capacity, a marginal or
contribution cost basis should be used rather than a full cost basis. Third, a differentiated
service characteristics premium (SCP), which is derived by comparing service characteristics
with average market competitors or a target competitor, is added into the adjusted pricing
basis. The term premium is used to indicate pricing above the market. [Arnold 89]
summarized their model as follows:

Pc(SL) = { VC + (PC * FC) + PG } * { 1 + SCP }

        (Ep) + (DURp) + V *
SCP =                       (Assuming equal weight of the premium factors)

Pc     = professional service price based on the extended cost approach
(SL) = price standard limits
VC     = variable costs
FC     = fixed cost
PC     = production capacity needed for delivering service
PG     = profit goal
SCP = service characteristics premium
Ep     = essentiality premium
DURp = durability premium
V*     = tangible intrinsic value-added

In contrast to traditional cost-oriented approaches, the major advantage of the extended cost-
oriented approach is that it incorporates premium factors into managerial profit-pricing
consideration. Also, the extended approach, to a certain extent, considers the competitive
advantages of product differentiation in service pricing. The major disadvantage is that the
service characteristics premium increases the complexity of the pricing task. Premium factors
are very difficult to evaluate objectively with in monetary terms. Also, it is reasonable to
believe that there should be more than the three premium factors shown in the model [Tung

Differentiation Premium Pricing Approach

[Arnold 89] proposes a premium service pricing approach which incorporates into the firm’s
pricing strategy recognition of the ability to differentiate the firm’s competitive advantages
from those of competitors. The differentiation premium (DP) comes from four factors:

  • availability premium (Ap)

  • reputation testability premium (Rtp)

  • commitment incentive premium (CIp)

  • price sensitivity premium (Psp)

Each of these factors ranges from +1 to –1. The service differentiation premium price (Pdp) is
equal to this differentiation premium plus an average competitors’ price (ACP). The
relationships can be shown as follows:

DP = f (Ap, Rtp, CIp, Psp)
Pdp(SL) = ( 1 + DP ) * (ACP)

Pdp = differentiation premium price after SL adjustment
DP     = differentiation premium
Ap     = availability premium
Rtp    = reputation testability premium
CIp = commitment incentive premium
Psp    = price sensitivity premium
ACP = average competitor’s price
(SL) = governmental or industrial price standard limit

According to [Arnold 89] availability refers to the number of services as well as the types of
services available to the consumers. Reputation testability describes the degree to which the
service performance can be evaluated objectively prior to consummation. Commitment
incentive refers to the relationship between profitability and the duration of the commitment
between the service provider and its customer. This relationship can have a major influence
on a pricing strategy. Price sensitivity depends on the number of service alternatives of which
consumers are aware. In general, the positive differentiation premium is associated with:

  • higher service availability

  • better reputation and less perceived risk of services

  • stronger commitment incentive of customer loyalty and long-term relationship

  • less price sensitivity resulting from service customization and product differentiation

In contrast to cost-oriented pricing strategies, a premium pricing strategy reflects market
competition. Price setting starts from competitors’ prices and is adjusted by a service

differentiation premium. However, this approach has some limitations. First, a service
premium is difficult to evaluate for two reasons: service factors should be more than the
above four factors such as high involvement/low involvement purchase, consumers’ service
knowledge in general, etc.; moreover, a differentiation premium is very abstract and
subjective. Second, the approach is more complex than the traditional cost-oriented method.
Third, using competitors’ prices to set a firm’s own price may not consider an organization’s
demand adequately at that particular price level [Tung 97].

Client-Driven Pricing Approach

[Ratza 93] presents a client driven model for service pricing. The model is based entirely on
clients’ response to price, namely the quantity of service used and the number of clients
gained or lost. The major advantages of this model are the consideration of the relationship
between market share (demand) and price and the maximization of short-term and long-term
profit. The major disadvantage of this model is that it is built on the economic assumption of
static equilibrium along with zero marginal cost and constant or linear consumption. Because
service firms compete in dynamic and changing environments, this approach is too simplistic.
Moreover, it is not easy for a firm to acquire complete chronological data on demand and
price for simulation purposes, especially for a service innovation or a newly offered service.
Besides, stabilizing market price is one of the major tasks of marketing managers, for a
frequently fluctuating pricing strategy may damage the company’s image [Tung 97].

Bundle Pricing Approach

According to [Guiltinan 87] bundling is the practice of marketing two or more products or
services in a single package for a special price. Reasons for service bundling may be cost
structures that consist of a high degree of cost sharing and a high ratio of fixed cost to variable
cost [Dearden 78], or when demand for a firm’s services is generally interdependent [Tung
97]. [Schmalensee 84] states that the advantage of pure bundling is its ability to reduce
effective buyer heterogeneity, while the advantage of unbundled sales is its ability to collect a
high price for each good from buyers who care very little for the other. Mixed bundling can
make use of both of these advantages by selling the bundle to a group of buyers with
accordingly reduced effective heterogeneity, while changing high mark-ups to those who are
mainly interested in only one of the goods. Operationally, a firm ultimately must determine a

specific price discount which is computed from the combined prices of two or more
individual service items. Thus, before a “package price” of bundled services can be formed,
the individual price of each service has to be formulated. This approach requires more
knowledge of specific costs, demand elasticity and cross-elasticity and reservation price
distributions. Although the bundle pricing approach is much more complicated than the
previous five pricing approaches, the use of bundle pricing appears to have been expanded
significantly in recent years, especially for consumer services [Guiltinan 87, Tung 97].

Dynamic Pricing: Auctions and Competitive Bidding

We have previously discussed in this work that consumers perceive a higher risk regarding
the purchase of services because they cannot actually estimate the true value of a service
before the actual service delivery. According to [Bichler 02] dynamic pricing mechanisms
such as auctions and competitive bidding can help find a price in cases where the true value is
not known by any single person and the individuals’ estimates may be highly imperfect. In an
auction, an object is offered by a bid-taker to two or more competitive bidders. The bidders’
bids indicate how much they are willing to pay for the object. According to [Bichler 02] any
well-defined set of rules for determining the terms of an exchange of something for money
can reasonably be characterized as an auction12. An auction clears when it commands an
allocation based on the bids it has received. The competitive process of bidding consolidates
the scattered information about the bidders’ valuations. There a several common types of
auction mechanisms in practice. The following list is based on [Buyya 01]:

      • English Auction (“first-price open cry”): all bidders are free to increase their bids
         exceeding others offers. When none of the bidders are willing to raise the price
         anymore, the auction ends, and the highest bidder wins the item at the price of his bid.
         In this model, the key issue is how much consumers decide to bid. Any consumer has a
         private value and a strategy for a series of bids as a function of his private value, prior
         estimation of other bidder’s valuations, and the past bids of others. In the case of
         English auctions, a consumer’s dominant strategy is to always bid a small amount
         “higher” than the current highest bid, and stop when his private value price is reached.

      • First-price sealed-bid Auction: each bidder submits one bid without knowing the others’
         bids. The highest bidder wins the item at the price of his bid. In this case a consumer’s

     For an overview of the numerous types of auctions please refer to [Sheizaf]

      bid strategy is a function of the private value and the prior beliefs of other bidders’
      valuations. A consumer’s best strategy is to bid less than his true valuation and still win
      the bid, depending on what the others bid.

   • Vickrey (Second-price sealed-bid) Auction: each bidder submits one bid without
      knowing the others’ bids. The highest bidder wins the item at the price of the second
      highest bidder.

   • Dutch Auction: the auctioneer starts with a high bid/price and continuously lowers the
      price until one of the bidders takes the item at the current price. It is similar to first-price
      sealed-bid auction because in both cases bid matters only if it is the highest and no
      relevant information is revealed during the auction process. The rate at which price is
      steadily reduced is determined by the auctioneer, who has a reverse price below which
      he will not sell. If the auction reduces the price to reverse price without finding a buyer,
      the auction terminates.

Auction mechanisms create an optimal outcome for both the seller and the buyer by ensuring
that prices match current market conditions. In other words, auctions are a way of determining
the actual demand situation. Traditionally, the high transaction costs have limited their
application to specific sectors such as finance, commodities, and art [Birchler 02]. However,
with the rise of electronic marketplace for all sorts of products and services – specifically for
business-to-business transactions – we expect a further shift away from fixed pricing and
towards dynamic pricing models.

4.1.5 Summary: A Framework to Grasp the Different Aspects of Pricing
As we have shown in the foregoing sections, pricing involves many different aspects whose
interrelations are rather difficult to grasp. The aim of this section is to present a framework
that allows one to master this complexity and be able to classify the pricing methods
previously presented according to selected criteria.

We have discussed the factors that influence price in detail in the previous section. In order to
gain an overview, one can distinguish between two sources of these factors: suppliers and

consumers. From the point of view of the price-maker, traditionally the supplier, a separation
between internal (organizational) and external (market) factors is useful.

The internal factors can then be broken down further into categories that stem from different
business aspects. On a top level, any organization defines its business strategy and goals, e.g.
profit growth or expansion of market share. These strategic decisions are authoritative and
must serve as an orientation for any business decision, including pricing. Furthermore,
financial aspects, i.e. cost, must be considered. As we have stated previously, prices must be
set so that in the long-run average costs are at least recovered. The companies overall product
or service portfolio must also be taken into consideration. Interdependencies among products
as well as life-cycles must be respected. From a marketing point-of-view, price must be
compatible with the marketing-mix, i.e. it must fit with product-, distribution-, and promotion
strategy. All these factors together ultimately make up the supply with which an organization
enters a market.

The demand which the suppliers face in a market is the aggregate behavior of the consumers.
Demand varies as the consumers react to changes in price, are influenced by the general
economic climate, anticipate other products and services, decide in the favor of substitute
products and so on. It is therefore impossible for a supplier to determine the exact demand at
any given point in time. However, there are pricing mechanisms and tools that allow price-
makers to estimate the actual demand. These estimates have to be taken into account when
pricing decisions are made. The extent to which consumers react to a change in price is called
price-elasticity which is a crucial factor that has to be considered by price-makers. The way
buyers act in a market depends strongly on the amount of information they possess. Sellers
must take into account how well their potential buyers are informed about their products and
their competitors. As we have described previously, the amount of information a consumer
possesses has an impact on his perceived risk and thus influences pricing possibilities. If
consumers have the possibility to easily compare prices between different suppliers,
competition and pricing are affected accordingly.

A further external factor is competition. Unless an organization is acting in a monopoly
market, it has to take into account the prices set by its competitors. Pricing policies such as

penetration pricing or price-skimming are oriented towards the competitor’s prices and serve
an organization to position itself in a market. Moreover, a seller may compare the quality or
value provided by its products and services with the respective products and services of its
competition. This may lead to setting a premium price, if the own quality provided is
estimated to be superior to the competition.

Pricing is the matching of all these factors with one another. This may happen “by nature” or
one may try to steer the pricing process by acting accordingly. Pricing “by nature” is to be
understood here in the sense of a market where all participants act as price-takers, i.e. they do
not individually influence price but accept the given market price as the produce of the
currently existing supply and demand. Of course, this is hardly ever the case in modern
markets. Instead, individuals – suppliers and buyers – try to influence price in any way they
can. Setting the right price, i.e. the price that best serves an organizations goals at a given
time, is a very difficult task and the numerous pricing methods available are an indicator for
the complexity one faces when trying to set a price that takes into account as many of the
factors described above as possible. Figure 4 depicts the pricing process from a supplier’s
point of view.

     Business          Product
     Strategy                      Pricing                                        Competitors



    ledge                          Pricing
                       Pro-        Methods


                 Figure 4: Factors Affecting Price from a Supplier’s Point of View

Each of the pricing approaches we have discussed previously takes into account some of the
internal and external factors listed above. Table 4 gives an overview of the presented pricing
approaches and their respective characteristics.

   Approach    Cost-      Competitive-   Extended   Differentiation   Client-   Bundle    Dynamic
               oriented   oriented       Cost-      Premium           driven    Pricing   Pricing
               Pricing    Pricing        oriented   Pricing           Pricing
  Criteria                               pricing

 Supply &
 Cost                                                                             ( )
                                                                                  ( )

 Attempts to


 Complexity      low          low          high          high         medium    medium      low

                  Table 4: Characteristics of Common Pricing Approaches

5     The Specialties of Pricing IT-Based Services

We will present the specific pricing models and methods used by ISPs, ASPs and in grid
computing in the seventh chapter of this work. However, there are some general specialties
concerning the pricing of IT-based services we shall describe in the following. By illustrating
the various market participants’ perspectives on pricing we shall obtain a comprehensive
understanding of the market environment, so that we can formulate the requirements on
pricing models and discuss their effects on the market. Moreover, we shall also discuss the
pricing of special services, such as those of dynamic virtual enterprises, and analyze the
impact of brokers and standards on service markets.

5.1    Lock-in
A specialty of purely digital services is that marginal cost for providing the services to one
additional customer is often next to zero. This allows service providers to attract new
customers by giving them access to their services “for free”. This will not yield any profits for
the provider, but [Shapiro 98] describes how service providers may attract new customers
with free offers and then attempt to create a lock-in situation, i.e. reducing the consumer’s
number of brand choices, increasing his switching cost and eventually making him dependent
on the provider’s services. Lock-in may occur through the use of proprietary technologies,
offering special value-added services and loyalty programs and so on. Once a customer is
locked-in, the provider may raise prices to regain the revenue he first waived by offering its
services free of charge. For the provider, the danger of proceeding in such a manner is that
consumers may only take use free services offered upfront, without ever turning into revenue-
paying customers.

5.2    Versioning
A major advantage of digital services is that they can easily be offered in different versions.
Versioning allows a service provider to offer the same service with different features and at
multiple service levels at different prices, thus enabling him to skim the willingness to pay
from different groups of consumers. Digital services may also be subject to customization, i.e.
allowing the consumer to configure a service exactly to his needs. If the provider can offer a

respective pricing scheme, he will again be able to maximize the profitability of the individual

5.3    Network Effects
A phenomenon that is found throughout the markets for IT-based services is called network
effects. This phenomenon was first described by Robert Metcalfe and is therefore also referred
to as Metcalfe’s Law. It describes how the value of membership in a network is enhanced by
an increase in the number of other members or in the other members’ usage of the network. A
classic example for network effects is the telephone service. As more and more people owned
telephones, the value of the network to the individual rose as one could reach ever more
participants. In industries characterized by economies of scale in supply, firms lower their
production costs by drawing more demand for their product, allowing them to produce more
of the product and thus benefit from the economies of scale. However in network industries,
organizations can lower their costs without having to lure customers away from their
competitors – by interconnecting with competitors (i.e., joining a network with the
competitor), thus allowing its customers to reap demand-side economies of scale as if its
competitors’ customers were its own [Aviram 03]. The impact of network effects on pricing is
two-fold. One could expect sellers to demand higher prices from customers that join the
network later, as they will benefit from a higher network value than the customers that had
joined the network earlier. However, what is seen in practice (e.g. in the mobile
communications market) is that new customers are attracted by massive rebates whereas
existing customers are only rewarded sparsely for their loyalty. While such behavior may be
adequate during the growth phase of a network, it is important that firms do not miss the right
point in time to change their pricing policy and start rewarding customer loyalty to avoid
running the risk of creating dissonances among established customers. Of course, network
effects may also cause lock-in situations from the customers’ view (as we have previously
described), thus allowing providers to indeed charge premium prices. However, firms must
well consider to which extent they may maximize their profits without upsetting and
eventually losing valuable customers.

5.4    Increased Perceived Risk
We have previously discussed that the perceived risk of the customer is higher when
purchasing a service than in the case of purchasing a product. This effect is often increased
when dealing with IT-based services. The reasons for this are the rapid changing technologies
that many IT-based services build up on. Especially in the case of long-time service contracts
involving relatively new technologies that may also require an upfront investment on the side
of the customer, the prospective customer’s willingness to pay can be narrowed by an
increased perception of risk and uncertainty regarding the future development in the specific
technology. This risk is also borne by the provider to a certain extent; however the provider
will most likely find itself in a weaker position during contract negotiations, since it is subject
to a higher pressure of having to close a deal. It is therefore reasonable that providers create a
pricing scheme which reduces the financial risk on the customer. This can be achieved by
charging relatively low entry prices and then later raising service fees to regain the previously
waived profits. The attainment of customer lock-in is therefore again a key factor for
implementing a successful pricing strategy.

5.5    Global Competition
The fact that the Internet facilitates global communication and enables worldwide cooperation
and competition has been emphasized countless times throughout literature. However, the
impact of global competition is especially strong on the providers of purely digital, IT-based
services. Because the provisioning of these services is often not bound to a certain physical
location, the number of competitors may rise quickly for any prosperous service thus
intensifying price competition. According to [Porter 80], a firm has three basic strategic
options; cost leadership, differentiation, or market segmentation. In the case of strong global
competition, a differentiation strategy may be the most promising in most cases. If a service
provider can manage to differentiate its services from the competitors’, e.g. by offering
certain local or region-specific specialties, it can avoid the often ruinous price battles and
instead charge a premium for its services.

A further specialty regarding the pricing of IT-based services is the question whether a flat-
rate or usage-based pricing model is more appropriate. In a flat-rate model, consumers pay a
fixed price allowing them discretionary use of the service. Usage-based pricing models charge

the consumers for the specific amount of services they consume. The usage as well as
advantages and disadvantages of the two models has been discussed extensively in literature
(e.g. in [Odlyzko 01]) and shall be picked up again in section

5.6    Further Considerations

5.6.1 Pricing for SLA-Based Service Provision
Service level agreements (SLAs) are an important tool to support the transition towards a
service-oriented IT organization. SLAs describe a provider’s services in detail, they define the
respective measurement metrics (so-called key performance indicators), and ultimately
provide a price at which each service is made available. SLAs are contracts between the
service provider and the customer, and serve as a guarantee for both sides. For the customer,
the SLA is a security that the ordered services will be delivered in the required quality. For
the provider on the other hand, SLAs allow a clearer communication about the provided
services and are a protection against unjustified service requirements.

With regard to pricing, the use of SLAs yields several issues and side effects, which we shall
discuss in this section. When looking at the pricing of SLA-based services, we have to
differentiate between SLAs that exist within corporations, i.e. between IT departments and
business functions, and SLAs that are part of contracts between service providers and external
customers. The reason for this is that the objectives of external and internal service providers
regarding the generation of revenue are mostly different. We have already discussed the
necessity of establishing IT within an organization as an internal service provider earlier in
this work. However, whereas external service providers aim at maximizing their profits,
internal service providers are mostly considered cost centers (see section 3.3.3) that have the
goal of recovering cost.

Whereas the aims of service pricing may differ, some effects caused by the use of SLA’s are
valid for both internal and external service providers. One such effect is that SLAs help
customers to better understand the services of the provider. When defining SLAs, the provider
must use a language and terms that is understandable by customers that do not have an IT or
other technical background. Otherwise customers will simply not purchase the provider’s

services (in the case of an external provider) or the lack of understanding of the services will
result in the business not accepting IT expenditures (in the case of an internal provider).
Hence, services are defined in a way that is customer-oriented, so that the customer does not
have to bother with technical details such as hardware configurations, but can rather order a
comprehensive service that fulfills his needs (an example of such a service provision is given
later in this section). From the pricing point of view, such services that are provided in a
black-box fashion reflect the concept of bundling. Thus, a combination of services is sold at a
single price. Regardless of whether a commercial service provider uses the bundling
technique to handle a consumer’s differentiated willingness to pay for multiple services or
whether an internal service provider bundles services to facilitate the internal communication
of its offerings, both external as well as internal service providers will need to be able to
determine the costs of their services before they can set a price. A commercial service
provider will then add a profit margin on top of total cost, whereas an internal provider
operating as a cost-center will simply pass the cost of delivering the service on to the

To illustrate further effects caused by the use of SLA-based service provision, we shall
present an example of an internal IT service provider. The environment in which the provider
operates is a large corporation, with several business functions and a distributed IT
organization. The IT branch considered in the following has decided to become more
customer-oriented, thus providing the business functions with a service catalogue that lists all
available services in a form that is understandable also for non-technicians. One service
offered is the installation and support of an IT workplace. Hence, if a customer within the
corporation orders this service, he or she will receive a personal computer with the installed
standard software, ready to use. The service offered by the IT department therefore includes
the installation of the hardware, software licensing fees, as well as helpdesk support. To
determine the price for this service, the IT department has calculated total cost by adding up
investment costs (hardware), software costs, support costs as well as overhead. The
investment costs are determined by calculating the depreciation of the personal computer.
Software costs are mainly given due to the acquisition of licenses from external vendors.
Total costs of support as well as overhead are calculated from historical data and extrapolated
according to business trends. These total costs are then divided by the forecasted amount of
workplace installations, resulting in the costs per user. These costs are then charged to the

internal customer in the form of a monthly subscription charge. In this example, the IT
department has decided to use fixed prices, therefore not considering the actual situation of
supply and demand to determine price. The reason for this is that fixed prices are more
transparent to the user and a dynamic pricing scheme would yield an excessive overhead.
Moreover, there is generally no scarcity of resources to provide the service, and should a
bottleneck ever occur, resources could be acquired from external sources. To ensure the
recovery of actual costs, the aggregated costs are reviewed every six months. In this process,
actual service consumption and forecast are reviewed and prices are adjusted if necessary.
User feedback is also used to assess whether the services described in the SLAs match the
actual requirements.

The price given in the SLA is strictly connected to the services described in that agreement.
Thus a customer knows exactly what he can (or cannot) expect for his spending. However, to
be truly customer-oriented, the IT department offers additional services that are not part of the
basic SLA, at an additional price. Hence, internal customers should not feel limited in their
choices by receiving strictly defined services, but should rather have the possibility to order
additional services to completely fulfill their needs – if they agree to bear the additional cost.
It is also a common practice to price these additional services higher than actual cost, to
passively influence the consumption of these services. For example, users may order a mobile
workstation instead of a standard desktop computer. As the support of mobile users causes
higher costs than the support of regular desktop users, it is in the interest of the IT department
to keep mobile users to a minimum. Thus, by pricing the service above cost at a high
premium charge, IT can try to steer the level of system heterogeneity it has to manage. The
internal pricing of services therefore also serves political interests within the corporation. Of
course, such issues quickly exceed the competence of the IT department, and a discussion of
such pricing policies must be held at the corporate level.

Apart from the price and the detailed service description, the SLA also includes key
performance indicators by which the service quality is measured. In the case of the workplace
installation service, the performance indicators are not aimed at the individual service
delivery, but rather measure the overall performance of the service. A key performance
indicator in this example could therefore state that 90 percent of all workstation installations
are performed within five days upon order entry. When SLAs are used between commercial

service providers and their customers, they usually include penalty fees that are inflicted upon
the provider in case the targets measured by the key performance indicators are not met.
These fees are based on the fact that customers are generally not willing to pay the full price
for a service that is not delivered in the promised quality. The payment of penalty fees from
the provider to the customer is equivalent to an actual price discount, thus having an impact
on the provider’s profit. The financial risk of penalty payments must therefore be quantified
and included into the pricing calculations to transfer the risk (at least partially) to the
customer. Ultimately however, receiving a price discount is not the actual desire of the
customer, but it is rather to receive the agreed upon quality so that the provided service
effectively supports the customer’s higher aims. Especially in the case of internal service
providers, penalty fees are only of limited use. They may well serve as means of motivation,
yet they will not reduce the actual cost caused by the service provision, which must somehow
be allocated within the corporation. Hence, if services are not delivered in the required
quality, the discussion between the provider and the customer should not be focused on
punishment, but the two parties should rather communicate and cooperate to find ways in
which quality can be improved. This is of course also valid for external service providers as
they are interested in effective customer retention and must therefore build up a cooperative
relationship with their customers that allows the open discussion of quality issues.

The introduction of SLAs does not only affect the provider but also has an impact on the
customer’s organization. In order to make effective use of the service provided according to
an established SLA, the customer’s organization must also possess a certain maturity level.
The customer must be able to integrate the high quality services provided according to SLAs
into its own processes; otherwise the efficiency gains achieved by strict application of SLAs
are lost. Hence the agreement to specific service levels may not only inflict consequences
upon the provider, but may also force the customer to reconsider organizational and
managerial aspects of its business. Whereas the costs of a possible realignment on the side of
the customer are not included in the price paid to the provider, a provider bidding for service
delivery must also understand the customer’s current maturity level, and the costs caused by
the required adjustments. Only then will the provider be able to provide and price its services
in a manner that is in line with customer’s actual situation.

It has been observed in practice that the adaptation of SLA-based service pricing in a
corporation reduces service consumption in the period following the introduction of the new
charging model. This can be explained by higher cost awareness among the internal
customers. However, experience shows that after a certain time period (usually a couple of
months), service consumption reaches the old level again. It is therefore not possible to
conclude that the introduction of SLA-based service pricing will reduce IT cost in the long
run. However, it can be said that SLAs help IT organizations to better communicate the value
they provide and the costs they cause towards other business functions. Financial controllers
(in the case of an internal provider) and customers (in the case of an external provider) will
have a greater acceptance of cost and price because SLAs that precisely describe services help
them to better understand the value they receive. One can look at acceptance as being an
aggregate of both performance and cost. Without SLAs, many customers (both internal as
well as external) have experienced performance by their IT service provider, and may well
have been satisfied with the quality. However, they may often have lacked the understanding
of the costs caused by the delivery of these services. With SLAs providing higher
transparency in regard to service description, controllers and customers can better understand
IT costs, hence raising the level of acceptance for IT expenditures.    Appropriateness of SLAs Within Smaller Corporations
During the previous discussion we have implicitly assumed the corporations we considered to
be of a larger size. The question therefore arises whether the use of service level agreements is
also feasible for smaller companies. Intuitively, we would say that the establishment in small
companies with only a handful of employees would resemble an administrative overkill and
that the conclusion of contracts in such an environment would seem rather odd. Additionally,
IT departments are often not explicitly separated in smaller companies, as the support of the
used IT infrastructure is handled by employees that also occupy business functions. However,
we believe that the establishment of SLAs can also make sense for small corporations, as
SLAs can be used to monitor and control the quality of internal IT. Whether this is required
depends on the extent to which business is dependent on IT. In a company where IT simply
consists of a small number of personal workstations used to perform regular administrative
tasks, SLAs may indeed be dispensable. Yet, if IT has a direct impact on quality or is even
part of the products that are produced, the introduction of internal SLAs is reasonable. This is
even more so if the company has established SLAs with external partners or customers, for

example in the case of a small Internet Service Provider. In such a case, internal SLAs are an
important tool to ensure that the guarantees regarding service quality that have been given
externally are transposed internally.

An exact estimate on the size of corporation at which the introduction of SLAs is regarded as
necessary cannot be given. However, when deciding on introducing internal SLAs or not, the
nature of the interface between business and IT is an important indicator. Hence, as long as
the company is of a size that allows simple ad-hoc communication between business and IT,
and IT is not a business critical matter, then SLAs are not a necessity. However, when a
company has reached a size where a certain anonymity among employees has emerged and
employees working in business functions no longer have a clear image of the services and
value that is provided by the IT department, the establishment of SLAs may well be
reasonable, yielding the benefits for business as well as IT that we have discussed in the
precedent section.

5.6.2 Pricing the Services of Dynamic Virtual Enterprises
According to [Lau 03] advances in information technology and communication networks have
enabled the emergence of collaboration through strategic partnerships between organizations
that transcend geographical boundaries. New and flexible organizational structures have been
formed in response to economic opportunities and challenges, thus enabling organizations to
participate simultaneously in various forms of virtual enterprises. [Mertens 98] defines a
virtual enterprise as a combination of legally independent enterprises and/or individuals that
provides a service based on a common business understanding.

Whereas static virtual enterprises are based on strategic long-term cooperation between a
fixed number of partners, dynamic virtual enterprises have a more temporary character and
can be defined as a set of business partners that are linked dynamically by a virtual market
place without forming fixed business relationships.

The unstable nature of dynamic virtual organizations does yield some challenges regarding
the pricing of their services. First of all, the prospective partners have to find one another on a

virtual marketplace and come to an agreement on the conditions under which each one of
them will provide services or service components. Part of this agreement will be the price at
which a participant is willing to cooperate with the other members of the virtual organization.
This price is determined by each participant individually and may be set according to the
common pricing approaches and models we have described previously in this work. It is well
possible that multiple prospective business partners may bid against each other, possibly in
the form of an online auction. The process of matching prospective business partners, i.e. the
founding of the virtual organization, may again be supported by a central directory service,
similar to the architecture we have discussed in the third chapter of this work. Once the virtual
organization has been installed and cooperation is in effect, the output, i.e. the service
provided by the organization, needs to be priced. If the dynamic virtual organization is to
offer services to third parties it will have to find some form of appearance under which it will
act on a market. Similarly, the participants of the virtual organization have to find a way to
determine one single price for their services, since a potential customer is very unlikely to
accept having to pay each member of the virtual organization separately. A possible approach
to solve this problem would be the introduction of a dedicated coordinator whose value-add to
the virtual organization would be the integration of the interests of the individual partners.
The coordinator would represent the virtual organization towards potential customers and
would also, amongst other coordinative tasks, calculate a single price out of the partners’
individual offerings. The price could again be determined according to one of the pricing
mechanisms previously described, with the coordinator ultimately adding a profit margin for
its own services.

5.6.3 Pricing Service Reservations
Reservations are widely used throughout all services industries. Customers can reserve hotel
rooms, concert tickets or tables in restaurants in advance. A reservation includes a specific
price at which a service can be consumed at a certain time. A reservation is therefore identical
to a European call option, known from financial markets. Specifically, a European call option
gives its holder the right but not the obligation to purchase a predetermined object at the
exercise price at a specific time [Quan 02].

For a consumer, the value of a reservation is two-fold. On the one hand, the consumer has a
guarantee that he will be able to make use of a certain service at a specific time, which is
valuable in the case of scarce resources. On the other hand, the consumer can avoid the risk of
rising prices by having a guaranteed price. In the case of falling prices however, the
reservation loses its value.

Why should a service provider offer its customers the possibility to make reservations? In a
monopoly market, a provider would certainly not have a strong incentive to allow
reservations. In the case of competition however, offering reservations resembles an added
value for customers and is therefore a mean of providing superior service. Moreover,
reservations can be used to optimize the utilization of resources and can serve as a tool for
price discrimination.

While the pricing mechanisms for financial options have been discussed in detail in economic
literature13, the prices for service reservations are not very transparent. In most service
industries, consumers are allowed to cancel their reservation up to a specific point in time
without any financial consequences. This yields a financial risk on the service provider,
because services suffer from perishability, i.e. if they are not consumed their potential value is
lost forever. Service providers must therefore calculate this risk into their prices. The price of
the reservation is not transparent, as it is included in the fixed price charged for a service.
Another major difference is that, in contrast to financial options, service reservations are
usually not traded in practice.

In the following, we assume that there is a market for the trading of reservations for IT-based
services. Of course, for reservations to become necessary, we must assume that IT resources
(e.g. bandwidth, storage or CPU-cycles) suffer from scarcity. For example, a reservation
could be used to reserve a certain amount of CPU-cycles at a specific time, and at a specific
price. The initial price for such a reservation would be set by the provider of the service, in
this case the owner of a data centre. In its calculations, the provider would need to include the
risk of the reservation holder not making use of the reservation and therefore inflicting costs
on the provider. The costs are actually opportunity cost, as the service provider will have

     For a concrete option pricing model, please refer to [Black 73]

reserved a certain amount of hardware (to be able to provide the service to the reservation
holder) that will remain unused and therefore not generating revenue if the reservation is not
cashed. However, including the whole risk into the price of the reservation would mean that
consumers would eventually pay as much for the reservation as for the service itself, hence
abolishing the consumers’ benefit of cancellation. The risk is therefore distributed among all
consumers and included in the price of the service. If reservations can be traded freely on a
market, the price of a reservation will certainly depend on the price of its underlying service.
Further factors that have an influence on the price of a reservation are price volatility of the
underlying service and the time remaining until expiration of the reservation14.

In order to improve its situation, a service provider may allow cancellations of reservations
without a financial penalty only up to a certain point in time, after which a part of the service
price is charged to the consumer, regardless of the actual consumption. Another tactic is to
offer resources that have become available because of cancelled reservations at lower prices.
Yet this tactic must be considered carefully as it yields the risk of cannibalizing regular prices.

5.6.4 Pricing Guaranteed Service Levels and Risk Assumption
As we have mentioned in previous sections of this work, Service Level Agreements (SLAs)
have become an important type of guarantee in services industries. [Kashyap 01] suggests that
a guarantee contains two typical elements: a service promise or pledge which expresses the
firm’s willingness to engage in behavior considered desirable by its customers and, second, a
compensation offer in case of service failure. SLA’s reflect these two components of a
guarantee as they specify precisely the service quality to which a service provider commits
and do also specify the penalties inflicted upon the provider in case of insufficient service

Guarantees serve service providers to signal service quality towards (potential) customers and
help reduce the perceived risk. In fact, by offering service guarantees and committing to
contractual penalties, service providers assume the risk of not being able to deliver the agreed
services in the respective quality.

     For a detailed discussion of such factors, please refer to the respective literature in financial science

We presume that in order to be competitive, service providers have to offer guarantees that
include penalty payments to the customer in case the agreed service levels are not achieved.
Such a guarantee inflicts a financial risk on the service provider and needs to be covered. It is
obvious that explicitly adding a surcharge onto a service’s price to cover the guarantee given
would not result in a positive feedback from the market. In its pricing strategy, a service
provider will therefore have to calculate the probability of being obliged to pay reparations as
well as the amount of the payments. The resulting quantified financial risk is then split among
all customers and added into the price of the service.

Specific adjustments of SLAs that may seem trivial to the customer can cause a substantial
increase in effort and expenditures on the side of the provider. An example for such an effect
is the adjustment of the guaranteed availability of a system from 99.9% to 99.99%. While the
customer may assume the increase by .09 percent to be realizable with only little extra effort,
the provider may well be forced to charge a substantially higher price due to its increased risk
of not being able to match the aggravated requirements. Guarantees can therefore also serve a
provider as a means of versioning. By offering different service levels at different prices, a
service provider can better manage its risks and is also able to skim the individual risk and
payment willingness of different consumers.

Yet another pricing strategy consists of allowing customers to waive service guarantees and
giving them a rebate on price. By setting the amount of the rebate below the value of the
waived risk, providers can obtain an extra profit that can again be used to cover the risks of
the guarantees that need to be given to more risk-sensitive customers.

5.6.5 Pricing for Mid- and Long-Term Contracts
A key element of any service contract is the duration of the business relationship on which
consumer and provider agree upon. Whether a mid- or long-term contract is established also
affects price, due to the different interests pursued by service providers and consumers.

When negotiating a contract, a consumer will – apart from bargaining for the lowest possible
price and other beneficial conditions – attempt to minimize his risks. The longer the time for

which a consumer subscribes to a service, the higher are the uncertainties and therefore the
risks the consumer faces. One such risk is the loss of flexibility in case of a general price
change, i.e. the danger of not being able to take advantage of decreased prices (e.g. caused by
technological developments) because of an obligation to prices that were set at the beginning
of a long-term contract. Another common risk for consumers engaging in a long-term service
contract is to become highly dependent on the service provider, e.g. by adjusting business
processes to the specific requirements of the provider. Such dependencies lead to increased
switching costs on the side of the consumer, thus limiting its options when the contract
expires. Switching costs are (non-monetary) costs that arise when a consumer chooses to
switch to another service or another service provider. In summary, consumers will try to
maintain their flexibility and keep as many options open as possible, thus favoring short- to
mid-term contracts and avoiding the risk of being locked-in in any form by the providers.

From the providers’ point of view, long-term contracts are more attractive because they allow
the binding and eventually lock-in of consumers to a specific service. Following the lock-in
scheme we have described previously in this work, providers may offer low entry prices to
consumers to lure them into signing a long-term contract. The longer the duration of the
contract, the wider is the range of possibilities the provider has to create a lock-in and
eventually being in a stronger position for renegotiations once the original contract expires.
However, it can be assumed that consumers will profoundly evaluate their risks prior to
engaging in a long-term contract, and providers will have to offer certain incentives to
compensate for the consumer’s increased risks. Price renegotiations can also be in the interest
of the service provider (e.g. if an increase of costs is possible) and a respective negotiation
schedule should therefore be part of the initial contract. To reduce price risk for both the
consumer and the provider, price may also be linked to a general price index or industrial
benchmark. According to [Macskasi 02], if long-term contracts have close substitutes in the
form of repeated short-term contracts, then long-term contracts must be priced at a discount.
This discount might offset the consumers’ inconvenience of being locked-in and hence might
be, in the end, beneficial to consumers.

5.6.6 The Impact of Standards on Service Markets
The standardization of services differs from the standardization of products in the way that
services are not physically tangible and do not (to the most part) contain physical parts that
can be standardized. The inner workings of a service are business processes that often
represent core competencies of the service provider and can therefore hardly be standardized.
Hence, the focus of service standardization must lie on the services’ interfaces, defining the
input that is needed by a service plus the format and scope of the output that can be expected.

[Sieber 99] differentiates between standards on a technical and semantic level and standards
that are concerned with business processes. Standardization on a technical and semantic level
(e.g. on the basis of XML) may help to increase efficiency whereas standards on a business
process or coordination level might restrain flexibility. Besides technical standardization,
standards may also be established on an organizational level (e.g. coordination services such
as brokers), which help to solve the problem of multiple specific interfaces through a more
efficient organization by providing an intermediary function.

According to [Sieber 99] the economic benefit of standardization is also enhanced through
network effects. Direct network effects lead to positive marginal utility, which is achieved
with a rising number of network members. Moreover, commonly used standards generate
indirect network effects, e.g. when more and more application solutions are based on XML
standards. Network effects lead to a quicker achievement of critical mass, defined as the
number of participants necessary for a market to operate efficiently. From the individual
organization’s or consumer’s point of view, critical mass describes the number of potential
partners that can be reached with the help of a specific service market. For example, a
supplier can reach more consumers when participating in a trading community. Critical mass
and network effects are often difficult to quantify but they help to explain why there might be
only a few service providers for each type of service worldwide [Sieber 99], a phenomenon
that is also described as a “winner-takes-all” market [Shapiro 98] .

Standards have an effect on consumer behavior because they reduce the consumers’
uncertainty and perceived risk about investing in a certain technology. Consumers will be
more confident about subscribing to services that are based on standardized technologies and

that are therefore expected to last. Moreover, industry-wide standards will make consumers
no longer dependent on proprietary solutions of specific service providers, thus reducing the
danger of consumer lock-in. Standards will therefore lead to an overall reduction of the
perceived risk on the side of the consumers, thus leading to an increased willingness to pay.

Service brokers are another group of market participants that profits strongly from the
introduction of standards in a market. Because brokers need to be able to compare the
providers’ services in order to match the consumers’ preferences with the right offering,
standards are a prerequisite if a service market is to host service brokers. Only if services are
standardized, or if there is a standardized format or language that can be used to describe the
features of the available services, service brokers will be able operate. However, a
standardized format alone is not a sufficient requirement for brokering services to be possible,
because it does not guarantee that brokers will have access to the data needed. Partnerships
among brokers and providers are therefore still required.

Service providers are also affected if a standard is established in a market. As described by
[Shapiro 98], standards tend to shift the focus of competition away from competition on
features towards competition on price. Hence, if services are completely standardized, i.e.
they become commodities, competition among service providers will be driven mainly by
price. However, as we have described at the beginning of this work, services are often in their
very nature much more customized than physical products; therefore the danger of
commoditization of services is also reduced. Moreover, service providers may escape the
danger of commoditization by offering special value-adding features that are based on
proprietary extensions, hence revoking the standardization to a certain degree. Whether a
service provider can survive in a market without adapting to the market’s dominant standards
depends strongly on its bargaining power. Several strategies and tactics for organizations
waging a standards battle are described in [Shapiro 98].

5.7    Summary
As we have seen in this chapter, the pricing of IT-based services involves a number of
specialties that stem from the very nature of these services. Electronic services can easily be
provided worldwide, thus increasing competition. Moreover, they are often subject to network
effects, meaning that they become more attractive as they are used by more and more
consumers. As electronic services are not bound to any physical representation, they are also
destined for versioning strategies, i.e. they can be offered in multiple, customized forms at
only little additional cost. Yet, their intangibility also increases the consumers’ perception of
risk regarding service consumption and service providers must therefore provide some
additional form of guarantee. One type of such a guarantee are SLAs as they provide a form
of insurance both to the provider as well as the consumer. Another type guarantee we have
discussed in this chapter are service reservations, which allow users to consume a service at a
given time in the future at a specific price. This however inflicts an additional risk on the
provider and must therefore be allocated into price. Furthermore, we have shown that
depending on the temporal length of a service contract, providers have more or less
possibilities of applying lock-in strategies, i.e. binding consumers to their offerings. Finally,
we have seen that standards can increase network effects, reduce the consumers’ perceived
risk, are an enabler for brokering services, and shift the providers’ focus of competition away
from features towards price.

6     Pricing in Practice

After having looked at various aspects of pricing in general and pricing for IT-based services
in particular in the previous chapters, we shall now focus our discussion on the practical
matters of pricing. First, we shall present various perspectives on service pricing from the
market participants’ point of view by illustrating their respective interests, aims, and concerns.
We then analyze the restrictions given by technology and administration on pricing, and in
turn define the technological and administrative requirements that need to be fulfilled in order
to support effective pricing models. This is followed by a discussion of the impact specific
pricing models have on the market participants’ behavior. We shall conclude the chapter by
establishing guidelines for selecting the right pricing model.

6.1    The Market Participants’ Perspectives on Service Pricing
In order to attain a comprehensive understanding of the players that influence price in service
markets, we shall now present various aspects and issues related to the pricing of services
from the market participants’ different perspectives. Whereas we have already studied the
consumers’ and the providers’ views in previous sections of this work, additional focus is put
on the perspectives of service brokers and the involved third parties.

6.1.1 From the Providers’ Perspective
We assume that profit maximization is the ultimate aim of any commercial service provider.
In a competitive market however, a provider will often have to waive certain short-term
revenues to achieve its long-time targets. This tactic has been described in this work when we
discussed lock-in strategies in section 5.1. It is crucial that service providers are able to
estimate a customer’s lifetime value to the business. If a provider can calculate that a
customer will (or is expected to) generate large profits during his long-term relationship with
the provider, it is plausible to give him a certain upfront discount that may indeed cut short-
time profits, but by winning a long-time customer maximize overall revenue.

By analyzing their cost structure, providers can come to the conclusion that not every
customer is indeed a profitable source of income. Depending on the nature of the respective
service industry it can be more profitable for a provider to focus on large enterprise customers
than offering its services to all. Apart from their advertising strategy service providers can
select the type of customers they want to attract by choosing the respective pricing schemes.
As we have described previously in this work, various forms of price discrimination are
available, e.g. offering lower prices to consumers of large service volumes. It is important that
a service provider is consistent with its pricing policy over time in order to avoid the risk of
creating cognitive dissonances among consumers and existing customers. If a change in
pricing policy is unavoidable, this must not be communicated solely through price but in a
more comprehensive manner to forestall customer irritation and damage of the provider’s
corporate reputation.

It is a common practice for providers in several service markets (e.g. the markets for mobile
communications or low budget air carriers) to create nontransparent price structures which
make it nearly impossible for consumers to compare prices of different services and
providers. The level of information and market transparency for the consumers is therefore
artificially decreased by the providers, strengthening the providers’ bargaining power. Such
tariffs can most frequently be observed in markets that are characterized by a low number of
sellers and high entrance cost for new competitors. In markets where new competitors can
enter at relatively low cost (e.g. the market for web hosting), it is much harder for an
individual provider to uphold a nontransparent pricing structure, because a new entrant can
gain a strong competitive advantage by offering simple pricing plans to consumers, thus
forcing the incumbent providers to adjust their pricing schemes.

With the advent of service brokers and intelligent agents market transparency has increased.
This yields risks and chances for service providers. On the one hand, there is the risk of
increased competition that is solely driven by price arguments and the danger that services
that are different in nature are compared and forced to compete against each other. On the
other hand, brokers open up new markets and distribution channels for service providers.
Service providers must therefore consider whether they want to cooperate with service
brokers or not. Chances are however, that providers will not have a free choice, because in
electronic markets, price comparers and service brokers may well gather the information they

need about a provider’s services without its explicit consent. The question is therefore
whether the providers want to be an earning part of the brokers’ business model or only a
part. While providers can aggravate the brokers’ and price agents’ work by technical means,
in the long-run, cooperating is the more promising strategy.

6.1.2 From the Brokers’ Perspective
In a service market, a broker represents an intermediating function that provides information
about providers, services, standards and contacts [Sieber 99]. Brokers are used in the
informational and selection phases of a service where they support the matching process of
buyers and suppliers. The value provided by a service broker is reflected in the cost savings
resulting from making use of a broker’s services compared to having buyers and suppliers
negotiate on their own. This value results from the different functions a broker performs in a
market, described by [Bailey 97] as follows:

  • Aggregation: Instead of each consumer having to negotiate individually with a service
     provider, and each provider having to negotiate terms and fill the orders of individual
     consumers, a service broker can aggregate the demand of many consumers or the
     services of many providers. Potential benefits include reducing transaction costs, taking
     advantage of economies of scale, and reducing asymmetries in the bargaining power of
     consumers and providers.

  • Trust: Brokers are able to prevent the parties involved in a market transaction from
     behaving in an opportunistic manner. Because of their long-term involvement in the
     market, brokers have high incentives to ensure that market transactions are completed,
     and that each party involved lives up to their end of the contract. Since the parties
     involved in the transaction may need to interact with the broker again in the future, even
     if they will never do business directly with one another again, the broker is in a better
     position to prevent opportunistic behavior, compared to other market participants.
     Moreover, brokers are typically involved in many more transactions than any individual
     consumer or provider, and as a result may benefit from economies of scale in their
     investments in transaction monitoring technology or they may be better able to insure
     against certain transaction failures.

• Facilitation: The exchange of information between organizations is costly, especially if
  the information exchanged is implicit or contextual knowledge that cannot be easily
  articulated. In such cases, a broker can facilitate the exchange of information by
  coordinating the process and translating the information that is sent between the
  negotiating parties. Brokers can also reduce the overall processing and coordination
  cost. Full connectivity in a disintermediated market with m buyers and n suppliers
  would require each of the m buyers to connect to all n suppliers, for a total number of
  nm connections. A broker optimizes this situation by collecting full information about
  the sellers with n enquiries and then pass that information to all buyers using m
  additional connections, for a total of n+m transactions. In a market with a high number
  of buyers and sellers, the difference between nm and n+m – and therefore the value-add
  by the broker – is significant.

• Matching: Brokers can accommodate the need of consumers to locate an appropriate
  service provider and vice versa by becoming a focal point for this match. Brokers can
  perform this task better than consumers and providers could on their own because they
  offer more efficient price discovery mechanisms and can acquire better knowledge of
  the respective market’s supply and demand characteristics. Moreover, by participating
  in transactions with multiple consumers, different providers and maybe even different
  industries, brokers can analyze consumer preferences across services, suppliers and
  industries. Brokers can use this kind of information to improve their ability to match
  consumers and providers, or they may offer this marketing information to providers.

                       Providers                             Consumers
                                          Service Broker

               Commodity Services                          Customized Services

           Figure 5: Brokers Provide Consumers with Customized Services

With the four functions described above still holding true, a modern view on the role of a
broker sees the broker as a “customizing aggregator” of commodities (depicted in Figure 5).
From this perspective, a broker acquires large quantities of commodity services and then
treats these services as components that are combined to produce customized services to the
individual consumer. By proceeding in this manner, service brokers can realize economies of
scale and at the same time offer services to consumers that could not be offered in the same
form by the providers themselves. However, this intermediary role also yields a potential for
conflicts for the provider. On the one hand, a broker will have to represent the customers’
interests towards the providers and bargain for low prices, possibly being tempted to play the
individual providers off against one another. On the other hand, the broker is heavily
dependent on the providers, because they produce the resources the broker needs in order to
be able to offer customized services to the consumers. To solve this dilemma, a broker has to
partner up with the individual service providers and establish a cooperation that yields
benefits for both sides. A broker can offer a provider a guaranteed minimum contract volume
plus access to otherwise unreachable customers, whereas the providers may – based on the
guaranteed service volume – offer special discount prices and tariffs to the broker. Provided
that the broker will forward the preferential pricing conditions it receives from the providers
to the consumers (a behavior to which the broker will be forced to, in a competitive market),
such partnerships between the broker and the individual providers will yield benefits – in the
form of customized services at low prices – to the consumers as well.

6.1.3 From the Third Parties’ Perspective
There are a number of market participants that are involved in business transactions without
actually selling or buying any product or service. These so-called third parties specialize in
the provision of supporting functions that enable or facilitate business transactions. Service
brokers that were described in section (section 3.4) fall into this category. This section
presents further types and functions of third parties that are relevant to electronic service

If buyers and sellers that are previously not acquainted are to do business with one another
over an electronic network, they are in need of a means of identification and authentication.
One method for achieving this is to introduce a trusted third party into the bilateral

relationship. The third party then acts as a certification authority (CA) that can vouch for a
party by issuing a certificate that identifies the party or attests that the party possesses a
necessary qualification or attribute [Froomkin 96]. Such certificates are also needed in Public
Key Infrastructures15 (PKI), where trusted third parties not only issue certificates that link
public keys to particular persons or organizations, but can also handle key management, i.e.
the generation and distribution of keys to the involved parties. Trusted third parties may also
time stamp or watermark documents to provide senders with authentic proof of time they
relayed specific information. Moreover, trusted third parties can function as safeguards to
ensure the privacy of a market’s participants by allowing transactions to take place
anonymously [De Laat 01].

Further functions that can be provided by third parties are reputation tracking as well as
virtual mediation and arbitration services [De Laat 01]. In reputation tracking, buyers and
sellers are asked to rate their trading counterpart. According to this feedback, market
participants are assigned a certain rating that will serve future trading partners to be better
able to judge their trustworthiness. The provision of virtual mediation and arbitration services
aims at the fast, facilitated and confidential settlement of disputes over electronically
negotiated deals and contracts. However, as there are no successful implementations of such
services yet, it shall not be further discussed at this point.

The described services are often provided in a cumulative fashion, i.e. one intermediary acting
as a trusted third party may offer multiple services. This should evoke certain considerations.
If business relationships are based on the trustworthiness of third parties, it is crucial that
these instances are subject to strict controlling and auditing regulations. Yet, even the best
auditing mechanisms will be worthless if a trusted third party is unable to survive
economically and has to close down its business, taking along all its accredited
trustworthiness. For such reasons, it has been proposed that trusted services should be
provided by government controlled instances that do not directly rely on economic profit.
However, further investigation on this political debate would exceed the scope of this work.

  For an introduction to PKI please refer to the vast amount of respective literature available. A good starting
point is provided by [PKI].

There are further functions that can be provided by third parties to support the trading of
electronic services that are not so much concerned with the provision of trusted instances but
can rather be regarded as an outtasking of certain value chain elements that do not resemble
core activities of a specific service provider. Examples for such services are billing systems
offered by third parties (that may also provide risk assumption) and the outtasking of

The involvement of third parties in service contracts does have an impact on price. In the case
of trusted third parties, both service provider and consumer may have to pay a certification
authority or alike a certain fee, thus increasing the overall price of a service. However, the
returned value in the form of improved trust and therefore reduced risk may well exceed the
price demanded by the third party. Intermediaries that offer the outtasking of specific value
chain elements will charge the service provider for their services. In the case of a third party
offering billing services, the fee demanded from the provider may be a fixed percentage of the
billed service volume [Glidewell 00]. The service provider will then allocate this additional
cost by calculating it into the price paid by the customer.

6.1.4 From the Consumers’ Perspective
Ultimately, the most important requirement a consumer possesses regarding the price of a
service is that price is equivalent to the value he receives from the service provider. Of course,
if a consumer receives value which he estimates to be even higher than the paid price the
consumer will be even better off. Both value and price are not limited to monetary means, but
also include factors such as search cost or additional personal value that is difficult to
quantify. It is crucial that a consumer perceives price to be fair and that he is not being
exploited in any form by the provider. For this to be possible, it is necessary that consumers
are able to comprehend the service providers’ pricing structures. Only if consumers are able to
obtain comprehensive market information and can compare prices, will they be able to choose
freely between the offerings of the providers.

There are, however, situations where consumers will not choose the service with the lowest
price. For services that cannot be considered commodities, consumers will often be willing to
pay a premium price, if their perceived value of a specific service is superior to competition.

Apart from measurable superior quality of service, the additional value consists for the most
part of the reduced risk perceived by the customer. Examples for elements which can reduce
the perceived risk on the side of the customer are premium brands, guarantees, and
transparent pricing structures. Another reason for consumers not paying the lowest price is
that search costs are high and consumers estimate the cost of finding the lowest offer higher
than the difference in price they could achieve. This is the case in markets in which providers
are able to maintain nontransparent pricing structures. However, with the rise of broker
services and intelligent agents, search costs for consumers have been drastically reduced and
improved their bargaining power.

Consumers will also try to keep their options concerning the selection of a specific service or
provider and will try to avoid any form of lock-in by the provider. However, in certain
situations (e.g. in monopoly markets) lock-in will not be fully avoidable. In such situations,
[Shapiro 98] advises consumers to negotiate for an upfront sweetener (a discount or any other
beneficial conditions) that is as big as possible. Another key factor for consumers to avoid
being exploited by a lock-in situation is to keep their switching cost low. This can be achieved
by avoiding investments in proprietary systems and standards, preventing the leakage of
corporate knowledge to the provider and keeping one’s organization and processes as
provider-independent as possible.

As important as transparent pricing structures is comprehensible billing. Unless consumers
prefer flat-rate charging (to be discussed in the third part of this work) they want to be able to
understand for which services they were charged and for what reason. Moreover, after having
consumed a certain service volume or having been in business relation with a specific
provider for a certain time period, consumers expect to be rewarded for their loyalty. This is
especially the case in markets in which providers give massive rebates to new customers. The
consumers’ chances of successfully bargaining for a loyalty bonus are the higher the lower
their switching costs are. If consumers can switch providers at low cost, their bargaining
power is increased and they can demand a bonus from their current provider for guaranteeing
him future revenue.

6.2    Real-World Restrictions for Theoretical Pricing Models and
We have presented various theoretical models and concepts for the pricing of services in this
work so far. In this section we shall now analyze their practicability. First, we will discuss
possible causes of problems regarding price formation. Later on, we shall specify the
requirements on IT-infrastructure and administration.

6.2.1 Causes of Problems
Numerous pricing models and strategies have been proposed in economic literature, yet only
few of them have been successfully applied in practice. One of the main reasons for this is
that theoretical models often require the input of variables that are very difficult, if not
impossible, to determine in practice. For example, the differentiation premium pricing
approach presented previously in this work demands the quantification of a number of
premium factors that distinguish a provider’s services from competition. In practice, such
variables can only be based on best guesses and will not fulfill the scientific ambitions that are
attached to many of the proposed models. We have also previously described in this work that
providers will always have to make various assumptions regarding the actual market they face
as well as the behavior of that market’s participants, i.e. the consumers and competitors. For
one thing, providers will never know the exact demand for their services, nor will they be able
to precisely forecast the consumers’ reaction to any changes they make to their pricing
strategy and policies. However, while the development of a pricing strategy will always
include many decisions based on intuition, theoretical models can serve as important strategic
guidelines and help compare different alternatives.

Whereas the issues described above are of a rather general nature and – up to a certain degree
– apply to any theoretical model that is to be applied in practice, there are further, more
concrete restrictions that apply to the pricing of IT-based services. These restrictions are
based on technical as well as on organizational problems.

Technical problems stem from the lack of technical realizations of theoretical concepts, often
caused by a lack of technical standards. Examples for this are the various reservation models

that were designed for the solution of network congestion. Whereas numerous models have
been published, a large-scale implementation of such a market model for network resources
has not yet been successfully accomplished. A major reason for this is that the standards and
network protocols that are required to guarantee the respective quality of service have not yet
been established. According to [McKnight 00] the lack of practicability is based to a large
part on the overhead that such dynamic pricing mechanisms (i.e. auctions) would create in a
network. It is feared that the overhead of the pricing mechanism would indeed aggravate
network congestion, and the overhead in the detailed accounting and billing for usage
required in such a complex system would make such pricing schemes unworkable. Moreover,
technical solutions that are needed to effectively measure complex service consumption and
quality are yet to be introduced.

The organizational issues stem from the lack of experience with modern organizational forms
such as dynamic virtual enterprises or service brokers. If the service and pricing models that
we have described in this work are to be realized, then ways must be found that allow the
pricing and accounting of compound services, i.e. services that were assembled by a service
broker from service components. Yet, it is one thing to determine a price for a service, but
allocating the earned revenues to all parties participating in the provision of the service is at
least just as complex. This does not only apply to services that were composed by a broker,
but also to services that are delivered over a network and therefore made use of various
services provided by many different providers until they reached the consumer. From a
business point of view, new pricing models only become interesting and realizable, if the
respective charging and accounting mechanisms can also be deployed. While this is to a
certain extent a technical issue, there are also organizational aspects that need to be solved to
enable inter-organizational charging and accounting of services. The high complexity of
billing systems found in the telephone industry today can be taken as a hint on the intricacy
that has to be mastered once more and more services are deployed on a pay-per-use or on-
demand basis.

Apart from the large-scale problems and restrictions described above, providers that want to
introduce innovative services along with the respective pricing schemes and mechanisms will
face many more practical problems. For example, in order to be able to price a specific
service, a provider must know the exact scope of the services he will be obliged to deliver. In

some cases, this is not trivial to determine. Such problems are typical for outsourcing
services, where service providers often lack exact information about the workload they will
have to handle. The solution to this problem is a process commonly referred to as due
diligence, in which the outsourcing provider closely examines the outsourcer’s infrastructure
and business processes in order to identify the exact scope of the service that will need to be
delivered. However, the problem that still remains is that once the outsourcing service has
been established, the outsourcing provider will be liable to detailed reporting about the
effective workload being handled, provided that the contract is based on a flexible, pay-per-
use pricing scheme.

6.2.2 Requirements on IT-Infrastructure and Administration
Consumers will demand high transparency regarding pricing schemes as well as charging
mechanisms. Hence, even though the actual calculations on which a provider determines its
prices may be highly complex, consumers need to be able to easily find out the price they will
be charged and, if there are multiple pricing options, which rate is best for them. Providers
must therefore maintain an infrastructure that is able to indicate the actual price of a specific
service at any given time. This requirement is aggravated by the providers’ demand for
dynamic pricing, i.e. being able to flexibly adjust prices to changing market conditions. Apart
from transparency in pricing, consumers will also expect clear accounting and charging for
the services they have (or have not) made use of. Unless a flat-rate model is applied, providers
are in need of an infrastructure that is able to display the customer’s current balance in real-
time. This information must also be directly accessible for the customer, which implies not
only high requirements on system availability but also on data protection.

In order to provide customers with transparent billing, service providers must be able to track
the effective service volume and service quality that has been delivered to every individual
customer. Whereas the respective measuring mechanisms may be easy to install for simple
services such as the provision of storage space, other services that also require the
measurement of quality attributes (e.g. the services provided by ASPs) are more difficult to
track and bill. Providers must therefore possess an infrastructure and maintain administrative
processes that are capable of measuring quantitative and qualitative performance indicators.

With respect to billing, service providers require systems that are capable of processing also
very small payments (so called micro-payments) in an efficient manner. Moreover, billing
systems must be interoperable, i.e. data exchange with other service providers must be
possible in order to be able to charge for composed services. If a service provider practices
price discrimination, this must also be supported by the billing system, i.e. the system must be
able to match the correct tariffs with the respective customer.

Dynamic pricing implies that providers are not only flexible in adjusting their prices, but are
also able to adjust their supply according to changes in demand. Service providers must
therefore possess an IT infrastructure that is highly scalable in order to be able to react to
altered demand for specific services.

In order to apply cost based pricing, a service provider must be able to determine the exact
cost of providing a specific service. While this is certainly a requirement that applies to any
serious business, it is of paramount importance to the providers of IT-based services, as they
act in an environment which is extremely dynamic. In electronic markets, prices can change
constantly as demand for specific services varies and service providers must be able to react
quickly. If decisions on pricing are to be made under strong temporal constraints the data on
which these decisions are based on must be available in the same real-time manner. Thus a
service provider is in need of accounting systems that allow the exact determination of the
total cost of a service at any given time.

Security is also an important requirement that is imposed on the service provider’s
infrastructure, both from the provider as well as the customer side. The providers must
establish a trusted platform that ensures the correct pricing and billing of all services and
fulfills all requirements commonly attributed to data security, namely privacy, authentication,
integrity and non-repudiation [Cannady 01]. As we have described in section 6.1.3, this can
be achieved by making use of the services of a trusted third party. From the provider’s point
of view, protection against fraud and misuse of services must also be ensured by the system.

A summary of the various requirements on a service provider’s IT-infrastructure and
administration is depicted in Table 5.

                         IT-Infrastructure                  Administration

       Requirements      •   Reliable                       • Provide transparent pricing and
                         •   Scalable                         billing
                         •   Trusted / Secure               • Efficient in the processing of
                         •   Interoperable                    micro-payments
                         •   Accessible                     • Able to provide real-time
                         •   Able to measure quantitative     accounting
                             and qualitative performance    • Interoperable
                             indicators                     • Able to support dynamic pricing

          Table 5: Pricing’s Requirements on IT-Infrastructure and Administration

6.3     Effects of Different Pricing Methods on the Market, and on
        Supplier and Consumer Behavior
So far, we have discussed a variety of pricing models and policies that are to be considered
when making pricing decisions. We now want to analyze the respective impact on the market,
as well as on the behavior of suppliers and consumers.

As we have discussed in section 4.1.3 price discrimination can be applied by suppliers in
various forms. However, if the discrimination is evident to consumers, they will attempt to
receive the benefits that are awarded to the group of consumers that is most privileged by the
supplier. Members of this group, on the other hand, may be tempted to resell their received
discount to consumers that do not qualify for a special offer. In economic literature, such
dealings are referred to as arbitrage. It is in the interest of suppliers to prevent arbitrage, as it
will interfere with the original intents of the attempted price discrimination. However, with
the advent of electronic markets, consumers have been given powerful communication tools
to form coalitions in order to coordinate their purchases. Yet, suppliers have at the same time
been provided with new means to individually address consumers and presenting them with a
singular offer. Hence, price discrimination is still a valuable method in electronic markets;
however it must be designed in a way that doesn’t allow the unauthorized transfer of benefits.

As we have discussed earlier in this work, suppliers may attempt to uphold nontransparent
pricing structures in a market. Such a policy may be successfully pursued if coordination
among the suppliers in a market is achieved or if one supplier possesses a monopolistic status.
Nontransparent pricing structures are in the interest of suppliers as they reduce the consumers’

level of information about the market, and thus increase the supplier’s bargaining power. We
see two possible effects such pricing policies may have on the market. On the one hand, it
becomes attractive for new market entrants to penetrate a market by offering simple pricing
structures, which may force incumbents to introduce transparent pricing as well. On the other
hand, markets that are characterized by nontransparent pricing are also destined to be entered
by service brokers. These brokers will take over an intermediating function as they attempt to
master the (artificial) complexity of the supplier’s pricing structures and present simple and
comparable prices to consumers. In the long run, suppliers may therefore experience great
difficulty in upholding artificially complex pricing structures. Exceptions are markets in
which extraordinary entry barriers exist, yet these barriers will not prevent brokers from
offering their services, thus challenging the market incumbents to review their pricing policy.

The advent of electronic markets has made auction style pricing more feasible for many
products. Due to the increased reach, more bidders can participate in an electronically
facilitated bidding process. Hence markets for products that allow electronic selling shift
away from traditional pricing towards dynamic pricing in which buyers and sellers determine
price at the time of transaction. Because auctions tend to allow suppliers to achieve higher
prices (or at least the price that reflects the highest valuation on the side of the consumers),
suppliers have a strong incentive to apply this type of pricing to as many products as possible.
Consumers, on the other hand, may not have such a high preference for auction pricing, as
they tend to suffer (in the popular English auctions) from “winner’s curse”, which describes
the effect that buyers typically overbid for the product [Wolski 03]. A discussion of consumer
behavior in auctions would exceed the scope of this work; however game theory has produced
a number of insights that are applicable to auctions and which have been extensively
discussed in literature16.

One of the key questions when pricing electronic services is whether services should be
charged according to a fixed price (flat-rate) or usage-based pricing scheme. In a widely
recognized experiment (Project “INDEX”), the University of California has studied the
effects of pricing on Internet user behavior [Altman 00]. The results of the study show that
demand for internet services is highly sensitive towards different pricing structures. In the
experiment, internet usage decreased drastically as flat-rate tariffs were replaced by per-
     A comprehensive starting point to game theory is provided by [Game Theory]

minute pricing, and vice versa. Moreover, in the experiment, users were very well able to
exploit the different pricing plans available to maximize their benefit, e.g. by purchasing extra
bandwidth only when required, instead of buying a fixed-price high bandwidth offer.
However, as we will argue in section, the large majority of users prefer flat-rate
pricing. We assume that this is due to the increased comfort such pricing plans provide, as
individuals are not hassled with having to keep track of their consumptions and the constant
query for the lowest available price. As illustrated by [Odlyzko 01], the general acceptance of
a (communication) service is always accompanied by flat-rate pricing. We thus conclude that
in general, the introduction of flat-rate pricing in a market will lead to an overall increase in

6.4    Selecting Pricing Models

6.4.1 Success Criteria for Pricing Models
In order to judge the success of a pricing model, one first has to define the targets according to
which success is to be measured. This measurement can be performed from different
perspectives, such as from the suppliers’, the consumers’ or from a general economic point of
view. We shall discuss the different aspects in the following.

From a corporation’s point of view, a pricing model is considered successful if it leads to the
achievement of pre-defined targets. According to [Wehrli 01] these aims that are pursued by
selecting a specific pricing strategy can be of corporate-internal or -external nature. Examples
for corporate-internal objectives are the alignment of production with capacity or the
achievement of a specific return on investment. External objectives may formulate targets
regarding customer acquisition, enhancement of market share or a specific strategic market
positioning. Of course, real-world pricing objectives will often include a combination of
internal and external targets.

Despite supporting the achievement of business objectives, a pricing model that is to be
successfully implemented must also master the organizational and technical requirements we
have discussed in section 6.2.2. Hence, it must not create an excessive overhead, both from a
technical point of view, as well as organizationally. Apart from the technical realization,

efficient handling of the accounting and billing associated with the model must also be

A key requirement for a pricing model to be effective in a market is that it is accepted by
consumers. [Wehrli 01] states that consumers expect honest prices. This honesty, which, for
example, is reflected in the form of adequacy or uniformity of prices, will lead to increased
confidence on the side of the consumers, thus strengthening a corporation’s customer
relationship. Also related to the consumers’ reception of price honesty is pricing transparency.
This means that consumers must be able to understand the pricing model before they buy and,
after purchase, must be presented with comprehensible billing for their consumption.
Moreover, consumer acceptance of a pricing model depends on whether consumers perceive
the pricing mechanism to be efficient. For example, an auctioning mechanism can be
considered economically efficient because a scarce resource is awarded to the consumer that
values the resource most. Yet, consumers may find the bidding process too cumbersome and
boycott the auction. However, depending on their market position and bargaining power,
suppliers may also be able to dictate a pricing model in a market, thus forcing consumer

The success of a pricing model can also be judged from an economic perspective. Success is
then reflected in the economic effectiveness that is achieved by applying a certain pricing
model. [Buyya 02a] presents a variety of criteria used for judging the effectiveness of a
pricing model, namely social welfare (i.e. the global good for all individuals), Pareto
efficiency (global perspective), individual rationality (better off by participating in
negotiation), and stability (pricing mechanisms cannot be manipulated and therefore behave
in the desired manner).

In summary, the success of a pricing model is the bigger the more of the above success
criteria from each perspective are satisfied. Hence, if a pricing model is able to allow a
corporation to achieve its strategic goals, is at the same time accepted by consumers, and
additionally fulfills the criteria of being economically efficient, the model can certainly be
considered successful. In the following section, we shall determine the criteria by which price
makers can select a potentially successful pricing model.

6.4.2 Choosing the Right Pricing Model
As we have stated in section 4.1.2 price is affected by a wide range of factors that stem from
different origins. It is therefore extremely difficult to define exact rules concerning which
pricing model to choose in a specific situation. Despite these limitations, we will in the
following attempt to provide general guidelines that shall support the selection of a feasible
pricing model.

When choosing a pricing model, one first has to determine the aims and targets that are to be
achieved with the respective pricing strategy. For example, focus may lie on gain of market
share, or one could attempt to establish a premium service at prices above the ones of
competition. Once the strategic targets have been set, a comprehensive understanding of the
market environment must be established. If one has to decide upon which pricing model shall
be implemented for a product, it is necessary to possess a profound understanding of the
relevant market, as some of the pricing models we have discussed require certain market
conditions to be fulfilled. For example, auctions are only effective if the good or service that
is to be auctioned off suffers from scarcity. Otherwise there will be an equilibrium price at
which more than one consumer can obtain the demanded product. Moreover, practical market
experience is also essential. For example, [Wehrli 01] states that price wars will emerge in
industries that are characterized by low innovation rates, relatively high fixed cost, and
excessive production capacities. In such situations, pricing policies should not be oriented
towards gain of market share, but rather focus on achieving positive marginal returns. When
selecting the optimal pricing model, such knowledge can be crucial.

Further criteria that are to be considered when selecting a pricing model are:

   • Business strategy

   • Costs (proportion of fixed and variable costs, marginal costs, transaction costs)

   • Market structure (monopoly, oligopoly, competitive market, etc.)

   • Predominance of other pricing models in market

   • Type of product (degree of commoditization)

   • Product characteristics (quality in comparison to competition, affinity to other products)

   • Availability of product

   • Available administrative support (accounting and billing)

   • Price transparency provided by model

   • Risks implied by the pricing model

   • Own bargaining power

   • Consumers’ bargaining power and pricing requirements

   • Existence of service brokers

With respect to the complexity of pricing discussed in precedent sections, this list may well
not be exhaustive. However, by considering these points, one can assess the pertinence of a
possible pricing model or policy.

We regard a detailed analysis of competition as crucial for the selection of an effective pricing
model. Hence determining the common pricing models in the relevant market is necessary.
The corporation that is evaluating a pricing model then must ask itself whether it pursues the
same aims as its competition, and, if so, whether it should adapt the same pricing model or if
the introduction of a novel model could lead to a favorable positioning in the market. If the
corporation determines that its targets differ from those of its competitors, it should consider
whether it has enough market power to establish a different pricing scheme. Therefore, it
should attempt to anticipate the reaction of its competitors, as well as those of consumers.

The selection of a pricing model should not be a static decision, but a corporation should
rather permanently assess the feasibility of its applied models as it reviews its pricing
strategy. This also implies that once an initial pricing model has been chosen, the supporting
organizational, technological and administrative environment must be setup in a way that
permits a later switch to a different model. Hence, a pricing model that implies very specific
organizational or technological installations may not be feasible for corporations acting in
highly dynamic environments.

The size of a corporation certainly also has its impact on the selection of a pricing model.
Smaller corporations may have the advantage of being quicker in adapting to changes in a
marketplace and may have better chances of positioning themselves in a niche market with a

specific pricing model aimed at a certain group of consumers. Larger corporations on the
other hand might have more flexibility regarding the design of their pricing models, as they
may possess better financial possibilities to subsidize aggressive (and possibly unprofitable)
pricing models for particular services. Hence smaller corporations must be much more careful
regarding the profitability of a chosen model, as their financial scope will most likely be very
limited. However their higher quickness to adapt to consumer requirements may well be

One of the key issues regarding the pricing of electronic services is the question whether
services should be priced according to a flat-rate or usage-based model. To illustrate the
debates that may evolve around the selection of a specific pricing model, we shall give this
discussion special attention in the following section.    Example: Flat-Rate vs. Usage-Based Pricing
Several works have been published discussing whether a service should be charged at a flat-
rate, i.e. at a fixed price allowing discretionary use, or usage-based. [Odlyzko 01] investigates
the history of various media of communication (e.g. mail, telegraph and telephone) and comes
to the conclusion that the success of a service was always accompanied by simple and
transparent pricing schemes. Odlyzko argues that conventional economic arguments for
quality and price differentiation are valid but that they conflict with some basic human
desires. Hence, a provider that offers flat-rate charging will more likely match the preferences
of consumers. Other arguments for a provider to use flat-rates are that they reduce the billing
overhead and therefore transaction costs and make the money flows more predictable.
Odlyzko further argues that flat-rates represent a form of bundling that allows the providers to
take advantage of uneven preferences among consumers for various service components of
the bundle, and thereby skimming the consumer surplus. Whether a service provider will
obtain more revenue from flat-rate or usage-based pricing depends on the distribution of
demands among consumers. According to [Fishburn 97] flat rate produces higher revenue for
the service provider for a majority of a large class of realistic demand distributions.

However, [Altmann 01] argues that flat-rate pricing is economically inefficient, because users
do not face the true marginal cost of usage, resulting in an over-usage and potentially higher

than socially optimal levels of infrastructure investment to meet the demand. According to
[Altmann 01] the high level of usage caused by flat-rate unlimited-access pricing plans have
the potential to reduce the overall network performance under broadband access technologies.
Moreover, light users subsidize heavy users.

In a formal approach, [Sundarajan 03] figures that in the presence of any positive transaction
costs (e.g. the costs for accounting and billing), sellers of information goods should offer their
customers a combination of usage-based pricing and unlimited-usage fixed-fee pricing. In the
absence of transaction costs, [Sundarajan 03] states that pure usage-based second-degree price
discrimination is optimal, even for information goods that are characterized by zero variable
cost. Furthermore, [Sundarajan 03] gives several managerial recommendations regarding the
pricing of services, which have been formally derived17. These include:

      • “If the administering of usage-based pricing involves a fixed per-customer transaction
         cost, which is triggered by any positive usage level, the optimal usage-based pricing
         schedule should include a minimum fee for usage up to a pre-specified level, and
         variable pricing beyond this.”

      • “Typically, the variable pricing described above will feature volume discounts.
         Moreover, the extent of discounting should decrease as marginal value increases, but
         should increase as variable transaction costs reduce, or as the customer distribution
         becomes less skewed.”

      • “In early-stage information markets characterized by a high-concentration of low-
         usage customers and a small fraction of active early adopters, low fixed-fee penetration
         pricing is a good strategy. This is especially true if there are setup or periodic
         infrastructure costs associated with administering usage-based pricing. As the market
         matures and the distribution of customers across different usage levels evens out, sellers
         should increase their fixed fees, and gradually expand their usage-based pricing

In practice, we can observe that most broadband Internet access services (e.g. DSL) are based
on flat-rate pricing schemes. Dial-up subscriptions are also offered at flat-rates by ISPs,

     For the formal derivations, please refer to [Sundarajan 03]

however in most countries – the big exception being the USA – users have to pay a
connection fee to their telecommunication provider, which renders the total cost of Internet
access variable. A study done by the University of Berkeley (project “INDEX”) has shown
that most users and ISPs prefer a flat-rate charge for basic services and a usage-based charge
when accessing a higher service quality that can be utilized on demand. The study also shows
that many users prefer flat-rate over usage-based pricing even if the fixed price of the flat-rate
is higher than the total cost of the accumulated usage-based charges [Altmann 01]. This can
be explained with the consumers’ demand for pricing simplicity and the reduced perceived
risk associated with flat-rate pricing.

6.5    Summary
While analyzing the market participants’ perspectives on pricing we have seen that there are
certain shared interests. For example, providers, consumers, as well as brokers are all
interested in the establishment of a trusted environment, and thus welcome the services of
third parties. Moreover, service brokers are beneficial to both service providers as well as
service consumers in the way that they aggregate consumer demand towards providers as well
as the providers’ services towards consumers and facilitate the bargaining process. However,
service brokers also render markets (i.e. services and their prices) more transparent, which is
certainly in the interest of consumers, but may conflict with the providers’ strategy of
maintaining nontransparent pricing structures. Service brokers thus find themselves in the
dilemma of representing the interests of consumers towards providers on which they are at the
same time dependent because they provide the resources on which the brokering services are
based on. We have therefore proposed a strategy for the broker that consists of teaming with
providers and establishing a cooperation that yields benefits for both sides. For example, a
broker may increase a provider’s reach, whereas the provider can offer special conditions to
the broker, which renders its services more attractive to consumers.

In section 6.2 we have analyzed the causes of problems that may occur when attempting to
implement a pricing model in practice. We have seen that in practice it is often difficult to
determine the parameters required by theoretical models, thus leading to a discrepancy
between the complexity of theoretical pricing models and the pricing techniques found in
practice. Moreover, we have identified technical problems, such as the lack of standards, but
also organizational issues, e.g. the excessive administrative overhead, that can render many

pricing models impractical. The most important requirements on hardware infrastructure we
have derived from this are reliability, trust, and interoperability. On the other hand,
administration must be able to provide transparent pricing and billing, perform real-time
accounting and support dynamic pricing mechanisms.

In section 6.3 we have noted that price discrimination strategies need to be reviewed in order
to be adapted to the networked economy. Moreover, we have shown that with the advent of
service brokers it has become more difficult for providers to maintain nontransparent pricing
structures. Furthermore, while the increased connectivity has rendered auctions practicable in
many new areas, providers must first consider whether consumers are truly willing to
participate in the bidding process.

In section 6.4 we have argued that the success of a pricing model can be judged from different
perspectives and figured that if a pricing model allows a corporation to achieve its strategic
goals, is at the same time accepted by consumers, and additionally fulfills the criteria of being
economically efficient, it can be considered successful. As we have seen, the selection of a
pricing model involves the consideration of a vast amount of factors, the most important of
which are competition as well as consumers. Moreover, we have noted that decisions
regarding the selection of a pricing model are never static, but must be reviewed constantly,
and that depending on its size, a corporation may have more or less flexibility in choosing a
pricing model.

7     Service Pricing for ISPs, ASPs, and in Grid Computing

In the seventh chapter of this work we shall discuss the theoretical and practical pricing
models available for ISPs, ASPs, and grid computing. Apart from the pricing models, we also
look at the providers’ architectures and illustrate organizational aspects. We shall also analyze
the differences between pricing various types of resources in grid computing, as well as the
dissimilarities of pricing in scientific and commercial grid environments. To conclude this
chapter, we shall discuss the differences between pricing for various types of providers.

7.1    Pricing the Services of ISPs

7.1.1 Overview of Theoretical Models
The pricing of Internet services has been discussed quite extensively in literature, starting in
the mid-1990’s, strongly driven by the explosive popularization of the Internet. In the
following overview, which is partially adapted from [Stiller 01], we shall present important
Internet pricing models that have been published.

Edge Pricing

The basic concept edge pricing is to charge the user only by the first network provider along a
data path that might make use of services from other providers. The charge to be paid includes
expenses for all different providers handling the respective data. Thus, multilateral contracts
are replaced by a sequence of bilateral ones, enormously reducing complexity and providing
user transparence. In the basic approach, the user defines the maximal total price he is willing
to pay as a sender or a receiver of data, respectively, as well as an upper bound for the
maximal number of hops. The charging information can be transmitted in the header of a
signaling protocol.

Volume-Based Schemes

Volume-based charging applies prices to the amount of data transmitted. This concept has
been applied commercially to X.25 networks as well as to different service classes of ATM

traffic. Of course, a suitable metering component is required to monitor the amount of data
transmitted. Most of these approaches use a scheme of price discounts based on several
thresholds. Stochastic network models are applied to derive price-functions that describe the
relationship between the current utilization of a (network) resource and the price to be paid for
using it.

Using Pricing Models to Solve Network Congestion

As we have stated previously in this work it is a general property of information that its
distribution to an additional consumer may occur at zero marginal cost. If one further assumes
that the investments made in the Internet’s infrastructure are indeed sunk costs and that there
is no scarcity-issue in information markets, one could come to the conclusion that the
provisioning of information via the Internet should occur at zero cost. Undermining this
theory is the fact that today many individuals do have access to various Internet services such
as the World Wide Web, e-mail and file transfers without being directly charged for their

To regard the Internet as one big source of free for all services is of course a misconception.
Apart from the costs that are borne by various commercial and non-commercial organizations,
resources on the Internet are also not unlimited. Today the Internet uses a “best-effort” service
model, meaning that network resources are allocated as effectively as possible, however no
quality of service guarantees can be made. When network congestion occurs, the users are
expected to lower their sending rate so that their collective data volume is equal to the
capacity of the congested point [Clark 97]. From the technical point of view, congestion
control on the Internet is handled by the TCP protocol. However, several proposals to use
pricing mechanism for solving Internet network congestion issues have been made.

Smart Market Pricing

[MacKie-Mason 97] argues that there are two different notions of efficiency – network
efficiency and economic efficiency – that can both be achieved by implementing a responsive
pricing. Network efficiency refers to the utilization of network resources, such as bandwidth.
Economic efficiency describes the relative valuations users attach to their network service. In
the Smart Market Pricing model presented by [MacKie-Mason 97] price is used as a form of

feedback to allocate resources under congestion. The model allows each user who sends
packages over a network to indicate how much he is willing to pay to get that packet onto the
network. The network gateway then sorts the incoming packets according to their “bids” and
admits to the network only as many as it can accommodate without degrading performance
below some previously specified bound. All users whose packets are admitted to the network
are then charged not the amount that they declared they were willing to pay, but the value of
the minimum bid on a packet that is admitted. Users therefore pay just the congestion cost
(the amount that the highest-valued denied packet would have paid for transport), and may
keep all of the excess value that they would have been ready to pay for delivery above the cut
off bid. This solution is therefore economically efficient and, by utilizing all available
network resources without causing congestion, network efficiency is achieved as well.

Expected Capacity Pricing

According to [Clark 97], the most significant service enhancement in the Internet would be
some means to distinguish and separately serve users with very different transfer objectives so
that each user could be better satisfied, and some means to limit worst-case behavior of each
user so that the resulting overall service would be more stable. [Clark 97] argues that marginal
cost in a network are non-zero only during network congestion and proposes the introduction
of a pricing model which ties the pricing of services to the expected capacity of a user.
Expected capacity allows a service provider to identify the amount of capacity that any
particular subscriber is to receive under congested conditions. Users then subscribe to a
certain profile that best reflects their expected usage. When a user sends a packet, traffic
meters at the network access points mark each packet as either in or out, depending on
whether the user has already used up the capacity he has subscribed for. As long as there is no
congestion, all packets are allowed onto the network. In case of congestion however, a
pushback notification is sent to the sender of out-packets, and out-packets are dropped. By
proceeding in this manner, both economic and network efficiency can be achieved. Because
all traffic – including out-packets – is allowed onto the network, network utilization is
maximized. Opportunistic behavior by individual users is suppressed, as available bandwidth
is split up among all users, according to the usage profile they purchased, thus leading to
economic efficiency.

Paris Metro Pricing

Paris Metro Pricing was proposed by [Odlyzko 98] and is based on subdividing the network
into different logical subnetworks, each of them handling packets on a best-effort base, but
charging different prices for them. The name of the model stems from the analogy to the
pricing scheme used in the Paris subway system, and it is expected that a more expensive
subnetwork will be frequented less often and is therefore able to deliver high-quality service,
however without formally guaranteeing quality of service.

7.1.2 ISPs’ System Architecture
The ISPs’ system architecture mainly consists of three sections; the access server, the
application servers, and the internal network. The ISP’s access server consists of a customer
interface, effectively a modem pool for dial-up users, and for permanently connected
customers (such as the large corporation). The application server consists of a series of
partitioned servers or separate servers dedicated to each application being offered by the ISP,
such as e-mail, web hosting or newsgroups. The ISP’s internal network depends on the
number of Points of Presence (PoPs) it has. Customers using dial-up connection prefer to
connect to a PoP in a local call zone so that timed charges for long distance tariffs can be
avoided. For this reason, to take advantage of economies of scale and to offer a uniform
service quality, the ISP may need to have a number of PoPs, each permanently connected by a
Permanent Virtual Circuit (PVC) leased from telecommunication providers [Kariyawasam

7.1.3 ISP Pricing in Practice
Despite the vast amount of publications discussing theoretical models for the pricing of
Internet services, the actual pricing performed by ISPs is in many ways more simplistic. The
reasons why the theoretical models are only partially applied have been discussed in the sixth
chapter of this work. How do ISPs determine the prices for their services in practice? Of
course, any ISP will have to set prices that are competitive in its relevant market. However,
assuming that the goal is to maximize profit or at least recover cost, an ISP cannot simply set
its prices according to competition, but must rather determine its cost and then use this

information to evaluate pricing strategies. Cost-based pricing is therefore the most convenient
pricing approach for ISPs in practice.

The costs for an ISP service are made up of two components: the specific costs than can be
directly assigned to the provision of the service plus an allocated share of the ISP’s costs of
basic operation. The latter include costs for leasing communication lines from
telecommunications providers, hardware costs for the hosting of user accounts (e-mail etc.),
software costs, support costs, customer acquisition costs, as well as general organizational
overhead costs. Table 6 gives an overview of these costs as well as the costs that need to be
considered when pricing other ISP services. Specific examples on how to determine these
costs are given by [Zigmont 99] and will not be discussed in detail here. However, it is
important that an ISP can accurately determine its total costs and is able to break these costs
down to a per-service as well as to a per-customer basis.

     Service    Dial-Up             Dedicated            Hosting          Basic Operation
                                    Access /             Services
     Costs      • PPP access        • Hardware           • Hardware       • Lease of Internet
                  cost,               (Routers)            (Servers)        backbone
                  depending on      • Leased lines       • Cost of          access line
                  number of           from tele-           band-width     • Hardware
                  users /             communication      • Domain         • Software
                  required            providers            registration   • Organizational
                  modems                                   fees             overhead
                                                                          • Customer
                                                                            acquisition costs
                                                                            / Marketing

                  Table 6: Cost Elements of the Services Provided by an ISP

Once the different costs have been determined, an ISP can calculate the prices of its offerings
by adding the target margin. Of course, the various pricing policies we have discussed in the
fourth chapter of this work are applied as well. Hence, services may be bundled or offered at
prices below actual cost to build up a strong customer base.

7.1.4 ISP Pricing Schemes Commonly Found in Practice
The main service offered by any Internet Service Provider is access to the Internet. According
to [Stiller 98] the pricing of this service consists of three basic elements: an access fee, a per-

call or connection/reservation-setup-fee and a usage-fee. The access fee, usually billed on a
monthly basis, is a fee charged for using an access link to the network. The price for the
access fee depends on the capacity, i.e. the bandwidth, of the provided link. The per-call or
connection/reservation-setup-fee is encountered either in connection-oriented networks or in
connection-less networks that implement reservation mechanisms for setting up connections.
Finally, the usage-fee to charge services on time-, volume- or QoS-basis can be used.

                                                       Content Pricing

                                                        Usage-based Pricing
                                     Related Pricing

                                                         Connection Pricing

                                                           Access Pricing

                        Figure 6: Components of Internet Pricing [Stiller 98]

The components of the prices charged by ISPs to their customers are depicted in Figure 6. The
figure also illustrates that independently of the basic transport service a content-fee can be
charged. The content-fee is usually omitted for general content such as publicly accessible
web pages or e-mail delivery, i.e. services which are considered to be included in the basic
Internet access services. However, special content – so called value-added services – can be
charged separately.

Pricing schemes applied by ISP’s must not necessarily include all of the components listed
above. Dial-up access to the Internet is usually provided for a fixed monthly fee, including
either unlimited use (so called flat-rate pricing plans) or a limited duration of connection,
with the time consumed above the threshold charged by the hour. Charging of fixed Internet
connections usually includes a fixed component depending on the bandwidth of the
connection and a variable component underlying the transmitted data volume. Another
popular variant of charging for Internet services is the so-called “bursty” rate method, where
the ISP periodically, e.g. every hour, measures the volume of data transferred over the
connection. For each charging period, e.g. every month, all samples are sorted by volume. A
fixed percentage, e.g. 5%, of the highest samples is discarded to eliminate unusual peaks, and
the highest remaining sample is used to define the bandwidth at which the connection is
charged [Stiller 01].

7.1.5 Pricing Between ISPs: Peering
The arrangement of traffic exchange between ISPs is referred to as peering. It is in the nature
of the Internet architecture that data traffic flows through the networks of multiple service
providers on its way from the sender to the receiver. Since the handling of the traffic inflicts
costs on the network owner, service providers have to find a way in which they can handle the
settlement of the exchange of traffic. It is common practice that larger ISPs which possess
their own backbone networks agree to allow traffic from other large ISPs in exchange for
traffic on their backbones. They also have an interest in exchanging traffic with smaller ISPs
in order expand their reach to regional end points. Mutual peering agreements are often
established without financial settlements. However, such agreements can be hindered by the
fact that the exchanged traffic is usually not balanced, i.e. one of the potential peers would
benefit more than the other. Peering may be considered counter-productive by one of the ISPs
if it figures that the potential peer would reap larger benefits, thus gaining a competitive
advantage at the first one’s expense. In cases where a difference in benefits is perceived,
peering parties therefore negotiate some form of compensation, e.g. one party paying the
other for the entire line between two networks, or the provision of free transit to a third
network by one of the potential peers [Stiller 01]. In order to agree on a peering contract, the
involved ISPs will have to assess the mutual benefits beforehand. According to [Stiller 01] the
metric that is used most often is the amount of traffic flowing in both directions, whereby
traffic sent from ISP A to ISP B is considered beneficial to ISP B, because presumably it
represents data that was originally requested by ISP B’s customers.

[Kariyawasam 01] describes different forms of settlement that may be introduced:

   •   Supplier-customer model: a unilateral charging model based on traffic delivered to
       downstream networks (inbound traffic). For this type of settlement, the traffic volume is
       measured and the receiver is billed in accordance with the volume of inbound traffic.
       Advantages of this arrangement include encouraging downstream ISPs to manage their
       network resources more efficiently and also encouraging the development of local
       content rather than downloading content from the upstream Transit Service Provider
       (TSP) or Internet Backbone Provider’s (IBP) network, where the original source of
       content is often from the USA. The disadvantage of the supplier-client model is that it
       tends to allow the supplier to free-ride on the local domestic ISP’s network by passing
       costs down to the ISP. In turn, the domestic ISP will try (although it may find it very

      difficult in an increasingly competitive market) to pass these costs on to the end-user. In
      effect, the supplier-client model ends up with the IBP or TSP setting the price floor
      faced by the ISP’s end-users.

   • Sender Keep All (SKA) approach: the interconnecting parties agree that traffic volumes
      are nearly balanced and it is more cost efficient to exchange traffic on a settlement free
      basis. SKA arrangements tend to work more efficiently where the interconnecting
      parties have networks that are roughly the same size, interconnect using similar capacity
      links, and exchange similar volumes of traffic. This type of arrangement describes the
      conventional view of peering on a settlement free basis described above.

   • Bilateral settlement: the interconnecting parties agree on a price for exchanging traffic,
      measure the difference in the bi-directional flow of traffic, and then reimburse the
      operator suffering the net inflow. This form of settlement is very similar to the
      international accounting rate settlement mechanism used for international accounting
      settlements between telephone operators.

   • Multilateral settlement: involves the joint funding of a Network Access Point (NAP),
      with each of the operators paying for linking to the NAP. At the NAP, the operators will
      exchange traffic on a SKA or settlement-based arrangement.

   • Discounted settlement: This form of settlement is a variation of the customer-supplier
      relationship described above, except that as the local ISP continues to download
      increased volumes of traffic from the upstream TSP or IBP, the local ISP will attract
      larger discounts. The advantages of discounted pricing is that cost savings are passed on
      to the ISP with increased traffic volumes, and these savings can in turn be passed on to
      the ISP’s end-users. In effect, as with the supplier-customer model, discounted
      settlements influence the retail price floor set by the domestic ISP, although in a more
      positive way.

Despite the different possibilities of settlement described above, peering with settlements is
very rare in practice. According to [Stiller 01] the only settlement scheme used in practice is
that one (smaller) ISP pays another (larger) one to connect to its network, but without the
right to transit to other networks.

7.2        Pricing the Services of ASPs

7.2.1 Overview of Theoretical Models
The services and the business models of Application Service Providers (ASPs) have never
enjoyed a large interest by the academic research community. Hence, the amount of
publications discussing theoretical pricing models for ASPs is very small. One of the few
works discussing ASP service pricing is [Cheng 02].

[Cheng 02] investigates the optimal pricing policies of web-enabled application services by
modeling the economic dynamics between the ASP and its potential customers. The model
considers a two-part pricing scheme, i.e. a scheme that consists both of a fixed as well as a
variable price component. It is also assumed that the ASP reimburses customers for the time
spent waiting for services due to congestions and that there is a minimal average performance
guarantee. With the detailed derivation omitted in this work18, [Cheng 02] determines a range
of optimal pricing policies, the form of which depending on the type of charging structure the
ASP chooses. These structures range from pure policies incorporating only a fixed or variable
cost component to policies employing fixed, variable and reimbursement price components.
Figure 7 summarizes these policies.

[Cheng 02] further concludes that with a linear cost function and uniform reservation price
distribution the optimal ASP profit is increasing at a diminishing rate. The optimal number of
subscribers reaches a limit as market potential increases, probably due to the performance
guarantee. Moreover, as the frequency of usage increases, the number of customers that can
be handled decreases, along with the ASP’s profit.

     Interested readers are referred to [Cheng 02]

Figure 7: Optimal Pricing Policies for ASPs [Cheng 02]

7.2.2 ASPs’ System Architecture
ASPs require an architecture that allows the distribution of software functionality over a
network such as the Internet. When we look at computing models for Internet-based
architectures, we can find that one of the key differentiators of the various models is the
location where the application code is executed. In the traditional desktop computing model,
applications are run locally on the client. In client-server computing, the execution of
programs is split up between the client and the server. The network computing model assumes
that software applications are dynamically downloaded from the network into the client for
execution [Furht 00]. Because in such an architecture clients require powerful hardware that is
able to execute programs, they are referred to as fat clients. In a server-based architecture on
the other hand, applications are executed on the server and clients need not execute the code
locally, hence they can be thin. Architectures for providing applications over the Internet can
be described using several tiers. An example of a four-tier architecture is depicted in Figure 8.

                     Presentation            Client 1         Client 2           Client n

                     Content                                    Web
                     Level                                     Server

                     Application                            Application
                     Level                                    Server

                     & Service     Billing System   Inventory System   Relational DB‘s   Other Systems

            Figure 8: Internet-Based Four-Tier Architecture for ASPs [Furht 00]

At the presentation level (the client tier), clients view Web content for information as well as
for application services. The second level (the Web server tier), consists of Web servers that
provide interactive views of information and support client-initiated transactions. Application
servers on the third level (the application server tier) are used to find the requested data and
services, make these available for viewing and carry out transactions. The fourth level (the
database tier) hosts a variety of data and services accessible by the application server [Furht

                                                 Data Input                               Application             Admin
        Web Browser                               Clients                                Access Clients           Client                                 Toolbox

                                                                                       Service Layer

                          Data Flow                                      App. State &                     Event Log    SW Licence
                         Management                                      Identity Mgmt                    & Tracking   Enforcement

                                                                                   Application Layer
        Management:                                                                                                                                       Support:
                         Service 1                                        Service 2        Service 3      Service 4    Service n
        Customer Care
                                                                                                                                                          Help Desk
                          Application Object 1

                                                 Application Object 2

                                                                                                                                 Application Object n
        Sales Support                                                                                                                                     Self Care
        Channel Mgmt                                                                                                                                    Error Tracking






                                                                                   Production Layer
                        App. Server                                     Mass Storage       Information      Disaster                                       Operations
         Databases                                                                                                         Data Output
                        Farm Mgmt                                          Mgmt             Life Cycle     Readiness                                      Log, Tracking

                     Figure 9: Architecture of an ASP Application [Furht 00]

[Furht 00] presents an architecture for applications provided by ASPs. The architecture is
depicted in Figure 9 but shall not be discussed in detail in this work. However, the
architecture features several components that are important regarding service pricing and
charging and which shall therefore be illustrated briefly. On the client side, the architecture
contains an administrative component, which gives customers the possibility of usage
tracking and reporting and allows billing presentment. These are important features, because –
as we have described in the sixth chapter of this work – transparent billing and charging is a
key customer requirement. On the side of the provider, the service layer features several
business management functions that enable automatic usage tracking and usage feed into the
billing solutions. Of course, the collected usage data is also important to the provider as a
means to control the effectiveness of the chosen pricing strategy, as well as a foundation for
the development of new pricing structures.

7.2.3 ASPs’ Cost Structure
According to [Jaruzelski 00] an ASP’s expenses can be grouped into four categories:

   • Solutions provisioning expenses are the costs of obtaining the legal rights to
      applications the ASP is planning to deploy. Typically, these costs include the software
      license fees an ASP must pay to independent software vendors (ISVs). Of course, if an
      ISV is providing ASP services directly to users, these costs will be marginal. However,
      there may be application development expenses related to making the application
      compatible to the delivery over a network. Specifically, the application code must be
      modified to enable delivery of the application to multiple users remotely and to allow
      the application to be partitioned on a data center server.

   • Solutions distribution expenses represent the costs of storing applications in a data
      center and delivering applications over a network. Specifically, these costs include the
      cost of servers, storage devices, and related software, bandwidth and the data center
      staff. ASPs that own a large amount of the required infrastructure themselves may be
      able to reduce these costs (on a per customer basis) by optimizing their infrastructure.
      However, these ASPs will face addition up-front capital investments associated with
      building their own data centers,

   • Service integration expenses include the costs of implementation, customization and
      configuration of the application, the customer’s legacy system and infrastructure. ASPs
      that possess enough expertise to perform these services themselves or have established a
      strategic partnership with system integrators may be able to buy these services at cost,
      hence creating a cost advantage over the competition.

   • Customer interface and administrative expenses include customer service and support,
      training, sales and marketing, research and development, and general and administrative
      costs faced by all ASPs.

According to [Jaruzelski 00] customer interface and administrative costs make up
approximately 40 percent of total cost. Service integration and solution distribution both
account for about a quarter of total expenditure, with solution provision (including software
license fees) making up the remaining 10 percent. Thus, the ASPs’ cost structure is dominated
by an excessive administrative and marketing overhead, whereas the ASP’s resource, i.e. the
software, is available at relative low cost.

7.2.4 ASP Pricing in Practice
ASPs offer a variety of pricing schemes under which corporate customers pay a periodic fee
(typically monthly) for software and services. Basic pricing and volume discounts are
determined by various measures, such as transactions, users, sites, or annual revenue. Several
pricing models have prevailed in the ASP market [Turisco 00], namely:

  • Per-user model: the per-user-model can involve either a flat-fee per user or a tiered fee
     structure based on the level of usage. The model is usually chosen by customers whose
     users all make frequent use of the ASP-delivered application. According to [Turisco 00]
     a tiered fee structure can be appropriate for a human resources ASP application, for
     example, where HR staff would need to access the application many times a day, but
     other employees would access it only rarely.

  • Per-transaction model: Under the per-transaction model, a rate structure is used to
     account for different automated processes of functions. Typically, customers are given
     discounts for various transaction volume levels.

  • Percentage-of-revenue model: The percentage-of-revenue (“slice-of-the-pie”) model is
     similar to the per-transaction model in the way that the monthly fee will fluctuate as the
     level of activity changes. However, the difference is that this model’s price calculations
     are tied to the amount of revenue the user generates each month. Yet, both models
     provide an added financial incentive to ASPs to keep the systems up and running
     because the ASP receives a higher fee when its services perform well.

  • Fixed-fee model: Under the fixed-fee model, customers pay a lump-sum monthly fee
     that covers products and services specified in the contract. This model is attractive to
     customers that posses a high risk awareness and want costs to be predictable. Variations
     of the fixed-fee models include separate fixed fees per site or per block of users. If other
     sites are added or the number of users exceeds the maximum listed in the contract,
     additional fees or a new fixed-fee rate structure are renegotiated.

  ASPs may charge additional fees, such as an application management fee that covers the
  cost of operating and maintaining the provided application and interfaces, hardware and
  the physical data center. Fees for support services, which may vary depending on the
  agreed service level, are also common. Additionally, ASPs demand fixed setup fees and
  variable charges for complex initial processes such as setting up multiple real-time
  interface for legacy systems and data conversion.

7.3        Pricing of Services in Grid Computing

7.3.1 Overview of Theoretical Models
Grid computing has been a widely discussed topic in academic research during the last couple
of years, leading to numerous publications. Within the research concerned with grid
computing, special interest has been devoted to the economic principles that are to be applied
within computational grids. The key question concerning the economics of grid computing is
how resources are to be allocated among grid users. It is generally assumed that the users that
make use of a computational grid are represented by (software) agents, which interact with
brokers to express their requirements, e.g. the budget they are willing to invest to solve a
problem, or a deadline (a timeframe by which they need results). Grid Service Providers
(GSPs) on the other hand, are in need of mechanisms that allow them to publish their services
and pricing schemes which help them maximize their profit and resource utilization [Buyya
01a]. To solve these problems, several economic models for resource allocation and price
setting mechanisms have been proposed and shall be presented in the following19.

Commodity Market Model

In the Commodity Market Model, GSPs (the resource owners) price their services
competitively. GSPs then publish their services in a business directory; however prices are not
made public. Pricing policies can be either flat or variable. In this model, a flat price means
that pricing is fixed for a certain time period and remains the same, irrespective of service
quality. In the variable case however, prices change as supply and demand changes. Pricing
schemes in the Commodity Market Model can be based on flat fees or actual usage. When a
user is in need of a specific service, he commands his agent to consult a service broker in
order to identify resource providers that are able to deliver the service according to the user’s
requirements. The broker then gathers suitable resources and establishes their prices by
querying the GSPs. The broker selects the provider(s) that best meet the user’s criteria and
sends their information to the user. The user then makes use of the providers’ services and
pays them as agreed.

     The overview is based on [Buyya 01]

Posted Price Model

The Posted Price Model functions in a way very similar to the Commodity Market Model,
except that it allows GSPs to advertise (post) special offers. Brokers then do not need to
negotiate directly with GSPs for price, but can use the posted prices, because they are usually
cheaper compared to regular prices. However, the posted-price offers will have usage
conditions, such as limitations to the timeframe within which the offered resources are to be
consumed. The establishment of transactions between users and providers is the same as in
the Commodity Market Model, however additional activities are performed. Additionally,
GSPs post their service offerings and conditions in a business directory. Brokers then check
whether any offers posted in this directory apply to the criteria given by the user before
negotiating with GSPs. If a posted offer matches the user’s requirements, the broker queries
the GSP for availability of the posted service.

Bargaining Model

In the previous two models, brokers didn’t negotiate prices but simply accepted the fees set by
the providers. In the Bargaining Model, brokers bargain with GSPs for lower access price and
longer usage duration. In this bargaining process, both brokers and GSPs have their own
objective functions and will negotiate with each other as long as their objectives are met.
Price negotiation will continue until either a mutually agreeable price is found or one of the
participants is no longer willing to negotiate. GSPs may eventually accept a lower price, if
that is the only chance of being able to sell their resources and not running the risk of low
utilization. The Bargaining Model is used by brokers and GSPs when market supply and
demand and service prices are not clearly established.

The three models presented so far, price is negotiated only between the user (or his broker,
respectively) and the provider; other consumers do not influence price. Therefore, the user has
no means of knowing how much others value the same service and must decide whether to
accept or reject the given offer depending solely on his private utility function.

Tender / Contract-Net Model

In the Tender / Contract Model the consumer requests bids for a specific service from several
GSPs and selects the one that best suits his requirements regarding price and timeframe. The
advantage of this model is that if one GSP is unable to deliver the desired service, the

consumer’s broker can search for other GSPs. The disadvantage of this model is that a task
might be awarded to a less capable GSP if a more capable GSP is busy at the time of
selection. A similar problem arises if a broker does not receive a response from a GSP at all.
Yet, the Tender / Contract Model allows specified contracts to be issued without negotiation.
The GSP that has been selected by the broker responds with an acceptance or refusal of
award. This can simplify the protocol and improve the efficiency of services.

Auction Model

In the Auction Model, negotiation is reduced to a single value, e.g. price. The difference
between the previously described models is that the Auction Model supports one-to-many
negotiation between one GSP (the seller) and multiple consumers (the potential buyers). In a
grid environment, auctions follow a specific procedure. First, the GSPs announce their
services and invite bids. Brokers then offer their bids and – depending on whether the auction
is open or closed bid – see what other consumers offer. The bidding process then continues
according to the various auction models we have described in the fourth chapter of this work
until only one bidder is left. Eventually, the GSP offers its service to the final bidder at the
agreed upon price.

According to [Wolski 03] English auctions are especially susceptible to “winner’s curse” in
which the buyer typically overbids for the good. In a grid environment, the problem with this
overpricing is that it may lead to inefficiency. Wolski argues that if currency can be
exchanged for resources, then excess currency expended is equivalent to lost resource time.
This will result in user colluding to keep prices low (possibly extinguishing supply) or simply
refusing to participate. In either case, users may be “encouraged” by the market to understate
the value of the resources because of the possibility of overpaying.

Bid-Based Proportional Resource Sharing Model

The Bid-based Proportional Resource Sharing Model differs from the other models described
so far in the way that the resource which is up for bid is not allocated completely to the
highest bidder, but rather split up among all bidders proportional to the different bid values.
For example, if two users bid for the same resource, and the first user bids 2 units of “grid
currency” and the second user bids 4 units of “grid currency”, then the first user is awarded

one third of the available resource, and the second user receives two thirds. This model is
useful when a resource is owned and shared by multiple individuals that are all assigned
certain amounts of credit to make use of the resource. The model can also be employed by
GSPs for offering QoS for higher price paying customers in a shared resource environment.

Community / Coalition / Bartering / Share Holders Model

In this type of model, all participants in the Grid are both service providers and consumers. In
such a peer-to-peer environment, resources aren’t assigned a specific price, but the access to
resources is rather determined based on the individuals’ contribution to the grid. Those who
are contributing resources to their peers are allowed access to the others’ resources when in
need. There can also be sophisticated mechanisms that calculate the share of resources a
contributor can obtain and allow a user to accumulate credit for future needs.

7.3.2 Grid Service Providers’ Architecture
Since grid computing technology is currently still strongly under development and Grid
Service Providers are yet to appear in large numbers, a typical architecture of a Grid Service
Provider cannot be given at this point. Instead, we shall have a look at a high-level grid
architecture presented by [Foster 03a]. For a comprehensive overview of existing grid
architectures and implementations, please refer to the respective literature20. Later in this
work (in section 9) we shall present a grid architecture that is specifically designed to support
multiple pricing models.






                    Figure 10: High-Level Layered Grid Architecture [Foster 03a]

     A comprehensive overview of grid computing is given by [Berman 03]

[Foster 03a] presents a high-level grid architecture (depicted in Figure 10) that organizes grid
components into layers:

   • The grid fabric layer provides the resources (computational resources, storage
      resources, etc.) to which shared access is mediated by grid protocols. Fabric
      components implement the local, resource-specific operations that occur on specific
      resources as a result of sharing operations at higher levels.

   • The connectivity layer defines core communications and authentication protocols
      required for grid-specific network transactions. Communication protocols enable the
      exchange of data between fabric layer resources.

   • The resource layer builds on connectivity layer communications and authentication
      protocols to define protocols for the secure negotiation, initiation, monitoring, control,
      accounting, and payment of sharing operations on individual resources. Resource layer
      implementations of these protocols call fabric layer functions to access and control local

   • The collective layer contains protocols and services that are global in nature and capture
      interactions across collections of resources. Services on this layer may include:
      directory services, discovery services, brokering services, accounting services, etc.

   • The application layer comprises the user applications that operate within a grid

This architecture identifies the principal functions that are required to enable resource sharing
within a grid environment. Due to its high-level abstraction, it does not place many
constraints on design and implementation. However, it shall serve us as a reference
framework when looking at concrete implementations of grid architectures presented in other

7.3.3 Organizational Aspects
Because large-scale markets that use grid technologies to exchange computational resources
are yet to be implemented, there is a lack of practical experience in the effective organization
of service pricing, accounting and charging for grid services. However, research concerned

with the future of grid technologies has come up with models and concepts of how a grid
economy could be organized.

We assume that in an open grid economy there will be numerous GSPs offering similar
services competitively, implementing the pricing models presented earlier. If we further
assume that individual consumers will make use of services provided by multiple GSPs, the
resulting administrative complexity is enormous. Apart from the problems of matching
consumers with providers and price setting, there need also be mechanisms that allow the
proper charging, accounting and payment of the consumed services. In order to be able to
cope with the complexity, [Barmouta 03] suggests that GSPs will not bill their users directly,
but that there should rather be a grid-wide bank which coordinates the payments between
providers and consumers.

[Barmouta 03] proposes a model called GridBank, a grid-wide accounting and payment
handling system. According to the model, both GSPs as well as consumers open an account
with GridBank. The user then submits his application processing requirements along with the
QoS requirements (i.e. deadline and budget) to the grid resource broker. The broker then
interacts with the GSPs’ grid trading service to establish the price of services and selects the
GSP that best matches the users’ requirements. The broker then submits user jobs to the GSP
along with details of its chargeable account ID in the GridBank. Consequently, the GSP
provides the service by executing the user job, and measures the resources that are consumed
during the execution phase. After completion of the job, the GSP charging module contacts
GridBank with a request to charge the user account, including a resource usage record.

The GridBank model also supports multiple payment strategies, namely “pay before use”,
“pay as you go”, and “pay after use” [Barmouta 03]. The first policy is most appropriate for
services that have a fixed price. A simple funds transfer protocol is designed to enable the
consumer to request a funds transfer with the confirmation sent to the GSP. The consumer
establishes a secure connection with the GridBank to provide details of his account and the
GSP’s account, as well as the amount to be charged and the URL of the GSP. GridBank then
performs the funds transfer and sends the confirmation to the specified URL of the GSP.

The second payment model, “pay as you go”, may be used to avoid having to establish a
trusted relationship between the consumer and the service provider. According to this
payment scheme, service consumers pay the service provider dynamically for CPU time or for
each computational result delivered, hence minimizing credit risks on both sides.

Finally, the “pay after use” model emulates the credit card payment model. If the definite
service charge is unknown before the delivery of the service, the service consumer sends a
payment order in the form of a digital check to the GSP. After computation has been
completed, the GSP calculates the total cost and forwards the check along with the resource
usage record to GridBank for processing. For multiple jobs this can also be done in batches.

7.3.4 Differences between the Pricing of Data, Computation,
       Communication, Applications and Knowledge in Grid Computing
In previous sections of this work, we have discussed the services enabled by grid computing,
as well as the pricing methods of Grid Service Providers. However, when we look at the
services and resources that are deployed through grid infrastructure, we notice that these
feature individual characteristics that need to be taken into account when making pricing
decisions. To analyze these differences, we will look at the following types of resources: data,
computation, communication, applications and knowledge.

In order to price a good or service, one has to know the factors that determine its value to the
consumer. In the case of data, the value lies both in the available volume as well as in the
promptness with which the data can be accessed. Computation is estimated the higher the
faster a specific computational problem can be solved. Applications have the most value when
they provide customized solutions to users. The value of knowledge strongly depends on its
applicability and its availability at a specific point in time. Finally, communication itself has
no value, if there is no valuable content to be transmitted. In a grid environment however,
value can only be exchanged if communication is possible, hence effective possibilities of
communication do have a value in combination with other services. From a pricing point of
view, this implies that the price of communication is always bundled into any other resource

Next, a potential resource provider in a grid has to determine in what type of unit the resource
shall be provided. Computation may best be measured in CPU cycles or the time a single job
occupies the provider’s CPU. The best measurement for data is in bytes, indicating the
amount of data stored using the provider’s infrastructure. In the case of applications,
measurement is more complicated. As we have seen in previous sections of this work, the
usage of applications can either be transaction based or it could be measured on the basis of
temporally limited licenses, i.e. allowing to make use of an application for a specified period
of time. To determine a reasonable unit for knowledge is the most difficult, because a
standardized measurement unit does not exist. Knowledge is generally understood as
information applied to achieve a goal, solve a problem or enact a decision [De Roure 03].
Thus, due its subjective character, knowledge is indeed difficult to measure. Yet to price the
provision of knowledge, providers could base their prices on queries made to their knowledge
databases. Another possibility is to sell subscriptions that allow discretionary access to a
knowledge database for a certain period of time. Communication is measured in the amount
of data transferred (i.e. in bytes) and is usually easy to measure.

A further difference between the five types of resources is their level of commoditization.
From the pricing point of view, being able to differentiate a service yields higher flexibility in
the design of prices and allows premium charging. The reason for this is that the provision of
services that are not commoditized usually requires a special kind of competency that cannot
easily be copied by others. Data storage, communication and computation services are all
commoditized to a rather high degree, because they are based on standardized, homogeneous
resources. In these cases, possibilities of differentiation lie in technical quality aspects, having
only a minor impact on pricing. In the case of applications and knowledge provision however,
providers have wider possibilities to differentiate themselves from competition because they
may be able to provide exclusive resources (specific knowledge) or services (specialized
applications) that cannot be found anywhere else.

Versioning – and thus the application of more differentiated pricing schemes – is possible for
all five types of resources. Communication may be provided at different prices, depending on
the required bandwidth. In the case of computational resources, premium prices may be
applied to give users the possibility of obtaining a higher priority for their submitted jobs.
Applications may feature different levels of functionality at varying prices. For access to

knowledge services, versioning might be realized by creating different categories of content,
with highly valuated content only available to consumers paying the premium surcharge. In
the case of data, a premium service might include dedicated infrastructure for data replication
of a single user, to provide optimized service performance.

According to the paradigm of service orientation, resources and services should be priced
based on the actual usage whenever possible. This is possible for the services that allow a
reasonable measurement of service consumption. As stated above, this can be done for all five
types of resources with the exception of knowledge, where a flat-rate pricing scheme is more
applicable. Another factor influencing the choice of the right pricing scheme is the time
criticality of a service, as this may lead to a highly volatile demand, which can be regulated by
implementing an auction based pricing scheme. Users would then indicate their need for a
service at a specific time through their bids. Time criticality is most apparent in the case of
computational and knowledge resources, with the former more likely being subject to scarcity

If we assume that in a grid computing environment the delivery of a service may be
distributed among multiple providers, it may not be clear which entity (or entities) actually
sets the price that is ultimately presented to the consumer. We therefore assume that in every
grid there exists a middle layer (e.g. a service broker) between the consumer and the various
providers that shields the consumer from having to deal with the complexity of distributing
tasks among multiple providers. Due to this transparency the consumer will also not notice
whether a service is priced by a single provider, or whether a middle layer will present a total
price that is actually an aggregated price of multiple parties participating in the service
provision. With regard to the five types of resources we have defined above, we can assume
that a distributed delivery of service is not applicable to all. Whereas computational problems
may well be distributed among multiple providers and data may be replicated at various sites
(possibly operated by different providers), knowledge and applications will most likely stem
from a single source. This is due to the previously mentioned fact that these services cannot
be commoditized to the same extent as the other two. The process of pricing knowledge and
application resources may therefore seem less complex, as fewer parties are involved
compared to the case of distributed data storage and computation. However, the previously
described concept of combining multiple services and resources of all five types into atomic,

integrated services gives rise to yet another scale of complexity in pricing, as the interests and
claims of various providers need to be incorporated as well.

7.3.5 Pricing in Scientific Grids vs. Pricing in Commercial Grids
For a long time, computational grid environments have been a subject of interest exclusively
to the scientific community, with only little appliance in commercial settings. However, with
the grid technology becoming more mature, the deployment of commercial solutions has set
in, intensifying the need for economic- and business models. Yet, while profitable pricing
models are a crucial part of any commercial grid application, profitability has only been of
minor interest in scientific environments. The reasons for this are basic differences that exist
between commercial and scientific settings and concern the motivation of pricing resources.
Whereas in commercial environments resources are priced to ultimately maximize the
providers’ profits, the motivation of pricing grid resources in scientific environments usually
lies in the efficient resource allocation.

According to [Buyya 02] resource management in scientific settings has often been based on
cost functions that are driven by system-centric parameters that enhance system throughput
and utilization rather than improving the utility of application processing. Resources are then
treated as if they all cost the same price and the results of all applications had the same value,
even though this may not be the case. [Buyya 02] argues that also in scientific grids, end users
do not want to pay the highest price but rather want to negotiate a particular price based on
demand, value, priority, and available budget.

Therefore, [Buyya 02] concludes that even if in the scientific community pricing aspects may
seem to be of minor importance, economic models are needed to help manage and evaluate
resource allocation to users who prefer to use economic driven schedulers to effectively
utilize their resource credits. Various economic models that are inline with the argumentation
of [Buyya 02] have been discussed in literature and have also been presented in this work (see
section 7.3.1).

The economical models of section 7.3.1 cannot be assigned exclusively to either scientific or
commercial grid environments, but can rather be applied in both domains. However,
depending on the environment in which they are used, the policies according to which the
grid’s participants act will differ. We assume that in a scientific grid, participating providers
do not aim at generating profits, but want to find the most efficient allocation of their
resources and possibly allocate their costs. Hence the aim is to make all potential resources
available to as many users as possible. Providers will therefore implement the respective
pricing policies into their systems, without regard for profitability. This means that as long as
resources are available so that all user requests can be served, services will be priced at-cost.
Only when resources become scarce, scientific Grid Service Providers may apply competitive
(auction-based) pricing models. However, the aim of doing so is not a financial profit, but
rather the achievement of economic efficiency by determining the user with the strongest
demand for a specific resource. According to [Buyya 02] the resource providers’ strategy
consists of obtaining the best possible return on their investment while trying to maximize
their resource utilization by offering a competitive service access cost in order to attract users.
This statement holds true for both commercial as well as scientific settings. However, the
difference is that in the former providers will apply pricing policies that are profit-oriented
also when resources are not scarce. Moreover, commercial Grid Service Providers will apply
pricing policies that are commonly found in competitive markets, i.e. they might be willing to
sell their resources at a price below cost in order to attract new customers or to better position
their services against competition.

In general, scientific and commercial pricing models do not interfere, because they are applied
in systems that do not interconnect. However, we can think of a scenario where the owners of
a large scientific grid decide to make part of their resources available to the public at
competitive prices. Despite all the technical issues and security related concerns involved, this
could also lead to problems regarding pricing. The main issue in such a scenario is that two
different pricing models are simultaneously applied to price the same resources. On the one
hand, members of the scientific community expect free access (or at-cost charges) to
resources, whereas on the other hand commercial pricing models are applied to charge
external users. This may be handled by applying strict user authentication mechanisms;
however in times of resource scarcity other problems arise. The question then is whether
resources should be awarded to the commercial user that pays a profit-generating price or

whether priority is given to the scientific community. Moreover, external customers that are
paying competitive prices might expect guaranteed service levels and priority treatment.
Hence, a scientific institution that decides to offer its grid resources to commercial users must
well consider to which extent such a change would have an impact on the grid’s original user
community. Ultimately, these questions of general grid usage policy must be settled at an
executive level.

7.4    Summary: Differences between Service Pricing for ISPs, ASPs
       and in Grid Computing
In the preceding sections, we have looked at various pricing models and mechanisms that can
be applied by ASPs, ISPs, and in grid computing. We shall now try to determine the
respective differences of pricing for these service domains.

Basically, the offerings of Internet Service Providers, Application Service Providers and
providers of grid computing services are very similar, as all three provide digital services that
are made available through electronic networks. Hence, one could conclude that the problem
of pricing the respective services is the same for all three types of providers. However, if we
again review the description of their services that we have discussed in the second chapter of
this work, take into account the factors affecting the pricing of (IT-based) services we have
analyzed in the fourth chapter, and look at the pricing models we have presented in this
chapter, we can derive that there are indeed some differences regarding the pricing of the
respective services.

When we first look at the relationships that exist between the providers and the customers for
each service domain, we notice differences regarding their durability. Due to the high level of
solution customization required, the relationships between ASPs and their customers tend to
be mid- or long-term. Compared to ASPs the services of ISPs are more commoditized and
therefore the binding between providers and customers are not as strong. Finally, we have
assumed that in grid environments the traded services are perfect commodities, thus the
relationships between service consumers and Grid Service Providers are ad-hoc and
terminated after completion of every transaction. These differences have an impact on the

pricing possibilities available to providers. The lock-in strategies we have discussed in section
5.1 are only applicable if the provider possesses some means to bind customers to its services.
Thus, the more commoditized the services are, the smaller are the possibilities of customer
lock-in. From this point of view, ASPs are in the best position to apply lock-in pricing
strategies, i.e. attracting customers with low prices, locking them in, and then gradually
raising prices. However, if we assume that the classical ASP model of providing applications
is replaced by a web services model as we have discussed earlier, this would indeed result in a
commoditization, thus decreasing the lock-in potential. If lock-in possibilities are slim or not
existent, and other means of service differentiation are lacking, competition between
providers will be driven strongly by low prices, resulting in lower profit margins.

Of course, the chances of lock-in are often anticipated by the consumers, thus increasing their
perceived risk. As we have seen in the fourth chapter of this work, increased risk results in a
lowered perceived value on the customer side, hence lowering the willingness to pay. Service
level agreements (as we have discussed in section 5.6.1) are one form of providing the
customers with a guarantee to lower their risks. We have assumed that guaranteed service
qualities, which are a prerequisite for the commoditization of a service, are given in the
described grid markets. However, in the case of ASPs and ISPs, SLAs are important tools in
the pricing process.

The level to which pricing can be automated also differs between ISPs, ASPs, and in grid
computing. We have proposed markets for computational resources in which pricing is
performed completely automated, i.e. without a need for human interaction, once the potential
user of a service has specified his requirements. We have specifically described this model in
detail for grid computing environments. To which extent pricing can also be automated for
ISPs and ASPs again depends on the degree of commoditization of their services. Automation
of the pricing process can be supported by the segmentation of services into components.
According to the broker model we have described in section 8 the pricing of the customized
services could then automatically be performed simultaneously to the service composition

A further difference between the three domains exists in the number of available pricing
models. As we have seen there are numerous theoretical pricing models for Internet services,
and much research is currently performed to develop adequate pricing mechanisms for grid
computing. Application Service Providers on the other hand have never been of much interest
to the scientific community; hence the number of available theoretical pricing models is very
small. However, as we have seen in the precedent sections, there is often a large discrepancy
between the complexity of the theoretical pricing models and the pricing schemes found in
practice. Hence, to which extent the lack of theoretical models has to be considered a
handicap for the Application Service Providers’ industry is questionable. Yet, we assume that
increased research activity will – in one form or another – have a positive impact on the
attached industry.

In the discussion so far, we have assumed the service providers to be commercial and
corporate-external. Yet, if we again return to the paradigm of service-oriented IT, we must
consider corporate-internal IT functions as internal service providers. Depending on corporate
policies, internal service providers may apply pricing models that are more or less similar to
the ones applied by commercial providers. If corporate policies dictate that prices of internal
service provision are to be set according to market conditions, the applied pricing models will
be more akin to the ones used by external providers. On the other hand, a corporation may
decide to have services provided at cost internally, thus simplifying the pricing process.
Likewise, political issues and considerations regarding corporate governance (as discussed in
section 5.6.1) may also have their impact on internal service pricing. The latter may include
methods that could not be applied in competitive market (e.g. the artificial raising of prices),
but require the monopolistic circumstances that are given within a corporation.

8     Pricing for Service Brokers

The paradigm of service oriented computing – as we have discussed in the third chapter of
this work – generally assumes a dynamic environment in which both software functionality as
well as business operations are provided by aggregating specialized components into
customized services. Service brokers play a central role in this model, as they collect the
requests of the service users, query the services offered by the providers, select the most
suitable service, possibly combine multiple services, and eventually present a customized
service to the user. In this chapter, we shall first review the impact of service brokers on price
and pricing, followed by the discussion of three generic application scenarios for service
brokers, each making use of a different technology.

8.1    The Impact of the Existence of Services Brokers and
       Maximum Connectivity on Price and Pricing
When the participants of a market are connected through a network that allows the exchange
of information between all parties – hence creating maximum connectivity – price will be
affected. In general, the enhanced possibilities of communication between market participants
will lead to high market transparency. Thus, an individual can much more easily attain
information about the goods and services being traded, the sellers, the buyers and the prices
being paid. This leverages the bargaining power of the consumers, as they not only are able to
easily compare the suppliers’ services but can also communicate amongst one another to
exchange their experiences and may unite to aggregate their interests towards suppliers. The
existence of service brokers further facilitates the consumers’ search for the lowest price, as
buyers no longer have to negotiate with a large number of suppliers, but can limit themselves
to communicating their needs and willingness to pay to a few service brokers. Hence, costs of
searching for the lowest price are reduced, which increases the consumers’ chances of
actually paying the lowest available price, thus leading to a lower average price. Moreover,
service brokers may also aggregate the demand of multiple consumers, enabling the
realization of economies of scale, which can lead to lower prices.

Generally better-informed market participants (consumers and competitors) force providers to
profoundly consider their competitors’ prices when making pricing decisions. Providers must
assume that their potential customers have complete knowledge about the available offerings.
Thus, if a provider plans to charge a price that is higher than that of the competition, the
provider must be able to justify this premium. This justification (e.g. premium quality) must
be conclusive to the consumers, because otherwise they will turn to the competitors. Another
tactic for providers facing well-informed consumers is to artificially decrease market
transparency by creating complicated pricing structures that make the comparison of different
providers more difficult or impossible. However, as we have described earlier in this work,
such pricing structures can only be successfully maintained as long as all competing providers
follow this strategy.

Whereas the existence of service brokers in a market may lead to generally lower prices for
commoditized goods and services, suppliers may still be able to achieve premium prices – if
they are able to provide a product that indeed differs positively from the competition. There
will always be consumers that are willing to pay an extra surcharge if it allows them to
receive additional value. The existence of maximum connectivity can therefore be
advantageous to providers of premium services and outstanding goods as knowledge about
their existence will spread more quickly among consumers.

A general statement on whether prices in a market will rise or fall with the introduction of
electronic brokers cannot be made because the effect on price depends on the pricing
mechanism implemented by the intermediary. Whereas brokers that search and compare the
static prices posted by suppliers may help consumers find the lowest priced offer – thus
lowering the average paid price – brokers that implement auction-style pricing mechanism
may well cause higher prices. According to [Lee 98], electronic auctions often enable sellers
to achieve higher prices compared to traditional auctions. Lee argues that in auctions there is a
statistically significant correlation between the number of bidders and the average price.
Hence, with electronic auctions generally allowing more participants than traditional auctions,
prices are expected to be higher, increasing the seller’s welfare. From a seller’s point of view,
the introduction of auction pricing for as many products as possible is therefore preferable.
However, auctions are based on the scarcity of a specific product, which is not the case for

commodities. Moreover, the overhead caused by the bidding process of an auction often
renders this method of price-finding impracticable.

In the following sections we shall present three examples of environments that make use of
service brokers, each based on a different architecture. The first example illustrates the
exchange of services within a corporation based on OMG’s21 CORBA22 model. Second, we
present an e-government environment implementing the .NET23 framework. Finally, a third
example shows the trading of software components with the help of an architecture based on
Jini24 technology. The examples shall not only serve us to further illustrate the paradigm of
service oriented computing, but shall primarily be used to examine the three architectures
abilities to support effective and efficient service pricing.

8.2     A Model of an Intra-Corporate Service Exchange using

8.2.1 Background: Agonic Workflow Management
In today’s dynamic economic environment, flexibility and quick adaptation to change are
essential abilities for any corporation in order to stay competitive. Corporations are in need of
architectures that allow ad-hoc, inter-organizational cooperation. Hence, business processes
and workflows are not only to be designed to function efficiently, but it must also be possible
to adjust and realign them swiftly. In order to achieve the required flexibility, workflows must
therefore be composed of modular components that can be combined on-the-fly to provide the
required services. Moreover, the introduction of intra-corporate pricing and charging
mechanisms based on real money – hence affecting the individual departments’ balances – is
required to guarantee efficient provision and usage of services.

[Geppert 97] presents a model for market-based workflow management. The model
implements the principle of agoric workflow management in which the activities contained in
a workflow are regarded as goods that are traded on an electronic market. Instances that

   Please refer to [OMG]
   Please refer to [CORBA]
   Please refer to [NET]
   Please refer to [Jini]

require a specific service then contact a service broker with their requirements. The broker
determines the most suitable provider by requesting bids from all eligible processing entities
(i.e. the providers). The providers calculate the actual cost and execution times of their
services and transmit them to the broker who in turn chooses the lowest bidder. Awarding the
lowest bidder with the execution of the service thus optimizes workflow execution with
respect to overall cost and execution time.

In order for agoric workflow management to be possible several prerequisites have to be met.
First of all, workflows must be decomposable, i.e. it must be possible to identify and separate
the activities that make up a workflow. Furthermore, a common language to specify the
required input and the provided output of each activity is required. Owners of activities must
be able to participate in a market, i.e. they must be able to advertise their services and bid for
activity execution [Geppert 97]. To coordinate requests, bids and transactions between service
users and providers, a central market making instance is required; this is the role of the service

In the following we present a high-level model of how services can be exchanged within a
corporation through an internal electronic market making use of components specified by the
CORBA framework. First, the different components of the model are presented, followed by a
specific example which illustrates the functioning of the proposed model.

8.2.2 Prerequisites
[Sairamesh 98] presents a model of electronic markets for trading computation and
information services. The so-called NetBazaar model makes use of the CORBA framework
and specifically implements the CORBA Trading Object Service to match users with
providers. Although NetBazaar does feature pricing policies, [Sairamesh 98] does not give
specific details on how effective pricing of the traded services is handled. Moreover,
NetBazaar assumes a market in which there are multiple providers for the same service,
whereas in the scenario of an intra-corporate exchange it is assumed that for each specific
activity there is only one providing instance. With regard to pricing, the existence of one sole
provider per service leads to a monopolistic situation, which yields the question of whether a
pricing mechanism is even required in this specific case (one could argue that a monopolist

can set price at will). However, it is assumed that despite the de facto monopoly of the
individual departments for their service domain, prices are not fixed but vary with changes in
demand due to the influence of corporate policy, governance and internal controlling. The
volatility of prices will also depend on whether the individual departments are run as cost or
as profit centers. Hence, in the case of cost centers, if demand for a specific service rises,
allowing the providing department to realize economies of scale, the reduced marginal cost
should be passed on the service user in the form of lowered price. For these reasons it is
assumed that price setting mechanisms are needed also in corporate internal markets.

Despite the shortcomings described above, we shall in the following adapt NetBazaar, the
open market model by [Sairamesh 98], to create a model for the intra-corporate exchange of
services. Moreover, we shall determine to which extent existing CORBA components can be
used to support pricing mechanisms, and what components would have to be additionally

First, we must define the type of products that will be exchanged on the virtual market. In the
example at hand, this will be elements of workflows, i.e. professional services. In the case of
an IT department, these can be services such as registration of a new user in an information
system or the handling of a helpdesk support ticket, whereas in the case of a human resources
department services may include the recruiting of a new employee or the writing of a
reference. Of course, these services may be composed of multiple processes performed inside
the respective department, however from the market’s perspective the services are regarded as
atomic elements.

Next, it is necessary to define a common data format in which the different services can be
specified. There are various data exchange and semantic representation formats available, e.g.
the Extensible Markup Language (XML) or the Resource Description Framework (RDF),
however the discussion of these is out of the scope of this work and in the following we shall
assume that there is an agreed upon standard used inside the corporation to describe the
available services.

8.2.3 CORBA Services Used to Realize the Trading System
The main trading component is CORBA’s Trading Object Service which provides a query
language for users to query information about service providers. Moreover, it provides a
yellow pages service in which providers publish their services. To find a required service the
users specify a base service type, a list of constraints on the attributes of the service (e.g.
maximum execution time, maximum cost, etc.) and a set of policies for the trading system to
execute on its behalf. CORBA’s Trading Object service provides the following [Sairamesh

  • An interface for providers to register and advertise information about their objects
       (services) and their properties.

  • A query language for users to gather information about service providers and their
       advertised services.

  • A distributed search and match service in a federated system of multiple trading sites.

  • A proxy service to import the legacy service objects into the trading system. The proxy
       can be used to access service objects from LDAP (Light Weight Directory Access
       Protocol) and DNS (Domain Name Service).

Further key services that are required to realize the desired trading system include the
following (the CORBA framework contains specifications for all of these services):

  • Naming Service: The naming service is used by the trading sites within an organization
       to advertise their interfaces and properties of the trading servers. Such properties can
       include the current workload on the server or the location of the server.

  • Notification Service: The notification and event services are used for asynchronous
       events related to service contracts and transaction management.

  • Transaction Services: The transaction management guarantees the ACID properties of
       transactions that take place in the trading system.

  • Security Service: The security service provides an overall security framework to realize
       access control, accountability, trust and other security related requirements in the
       trading system.

8.2.4 Support of Pricing by CORBA
The CORBA framework does not specifically provide a service dedicated to the realization of
pricing mechanisms. The CORBA Licensing Service can be used to track the usage of
components and services; however it primarily serves as an accounting tool and can at best be
used to support pricing. Additional components must therefore be designed if dynamic pricing
models are to be implemented in CORBA based architectures. For example, an auctioning
service could be added to the trading system to allow bidding for service delivery in case of
scarce resources on the side of the provider. The auctioning service would then function as an
auctioneer, coordinating the bids of the service requestors. Moreover, the number of requests
for a specific service can be taken as an indicator for demand and could be communicated to
the service provider which could in turn use this information for resource planning and the
design of pricing schemes. In the example at hand, the trading system functions within a
closed environment (the corporation), i.e. the potential users of the system are known. It is
therefore possible for the service providers to denote specific pricing policies for individual
customers, e.g. based on special conditions that have been negotiated between two business
departments. This enhanced pricing information can then also be stored by the trading object
service’s yellow pages function.

8.2.5 Functioning of the Proposed Model
In this section we present a concrete example to illustrate the functioning of the proposed
trading system. We shall assume that a large bank A organizes the intra-corporate service
provision between the various business supporting departments through an electronic market.
All departments therefore specify their services in a standardized format and advertise their
offerings on the trading system. These specifications include, amongst other information, the
content and format of the output, price and the estimated duration until completion of the
service. For example, the human resources department might publish a service named hiring
which ultimately delivers a new employee to the requesting department. After publication to
the trading system (1), the specification of the service is stored in the yellow pages function of
the object trading service of the trading system. If, for example, the IT department of the bank
requires a new employee it will query the trading system for a hiring service (2). The trading
system will then contact the resource broker of the human resources department and enquire
whether the service can currently be delivered according to the published conditions or if
another pricing policy should be applied due to increased demand for the service or a special

agreement between the two involved departments (3). The trading system then informs the
service requestor, in this case the IT department, of the current conditions (4). If the IT
department decides to make use of the human resources department’s service, the trading
system will register the trade and inform the accounting system (5). To invoke the service, the
IT department connects to the human resources department’s object resource broker and
provides the required input (6).


                                    Trading System

      Dept                                                                            1              Dept
              ORB                    Trading                                                 ORB
      1                                              Naming      Notification                           1
                                     Object                                           3
                                                     Service      Service

      Dept                                                                                           Dept


              ORB                                                                            ORB
      …                                                                                                …
                                    Transaction      Security         Pricing
                                      Service        Service          Service
      Dept          2                                                                                Dept
              ORB                                                                            ORB
      n                 4                                                                               n
                                               Provider Information
      Service Requestors                                                                  Service Providers


                                               Accounting System

                            Figure 11: Trading Architecture Based on CORBA

8.3     A Model of an E-Government Portal Using .NET

8.3.1 Web Services as an Enabler for E-Government Solutions
According to [Spahni 02] governments should move away from their traditional hierarchical
organizational structures and become more service-oriented network organizations. In a
network organization, governmental authorities are in contractual relationships with service
providers which can be either other authorities or private institutions. Within such a network
of providers, services are exchanged across both organizational as well technological
boundaries. [Spahni 02] therefore argues that the key to success lies in the integration of the
systems of the providers and partners involved in the value chain of the governmental service
provision. The use of web services (as provided by the .NET framework) supports the

development of such network structures in the governmental domain by allowing the
integration of multiple governmental information systems and thus enabling more efficient
cooperation among authorities. Moreover, the creation of an integrated e-government
architecture based on web services not only leads to increased efficiency, but also allows
authorities to obtain and resell services from third parties (again, governmental or private),
which is especially attractive for smaller municipalities which could not offer certain services
by their own means. If web services can further enable the realization of an integrated service
platform instead of multiple stand-alone solutions, citizens will be more likely to access the
electronically provided services, thus allowing the forward integration of many tasks, which
in turn reduces the workload of governmental offices and leads to cost savings.

The basic concept behind the idea of an e-government service broker is to provide individual
citizens as well as organizations that require services from government authorities with a
single point of contact for all their needs regarding governmental services. Users can then
provide the broker with their “problems” (e.g. relocation to another city) and the broker will
arrange all required services with the respective authorities. From the users’ point of view, the
governments services will appear as a single entity, customized to solve their current problem.
Hence, the inner workings of the government, the cooperation between multiple authorities,
are transparent to the users who therefore do not need to know the organizational structures
and responsibilities of the different agencies. Citizens therefore find information more quickly
and redundancies, i.e. the same information stored at multiple sources of information (thus
risking inconsistencies) can be reduced.

8.3.2 Pricing of E-Government Services
With the increased cost pressure and service orientation the pricing of services has also
become an important issue for governmental authorities. However, the pricing of
governmental services differs from the pricing of commercially provided services in some
aspects. First of all, most governmental institutions do not have the aim of generating an
economic profit but rather target at the – at least partial – recovery of cost. Additionally,
governments provide a number of services where full cost recovery would inflict an excessive
financial burden on individual citizens and therefore cannot be priced to recover full cost, but

are subsidized by the state. Moreover, public institutions must also ensure that their common
services are available to all citizens, thus often not allowing a service charge.

The Public Management Service group of the OECD25 has established best practice guidelines
for user charging for government services. [PUMA 98] states that wherever relevant, pricing
of government services should be based on competitive market prices. In other cases, pricing
should be based on the principle of full cost recovery unless there is a clear rationale for less
than full cost recovery (as described above). According to [PUMA 98] consideration should
be given to differentiated prices for peak and off-peak periods as well as to offering priority
services at a premium charge in order to spread demand for services. Moreover, the
application of price discrimination should also be evaluated in the case of lower-income
individuals, users located in remote areas and large volume users of services. To which extent
these pricing considerations can be reflected in the proposed e-government portal shall be
discussed in a later section.

8.3.3 An Example E-Government Service Broker
To analyze to which extent pricing mechanisms can be implemented in the case of .NET
based web services, we present an exemplary e-government service broker architecture. In
this architecture, a service broker functions as a link between the various governmental
agencies and the citizens. Web services are used to make the services of the agencies
available to the public. The agencies on their part have integrated their various information
systems and databases with the help of the .NET framework and expose their provided
functionality in the form of web services. The e-government broker provides a portal website
on which the citizens can navigate with their web browsers and provide information about
their needs. The broker then interacts with the agencies’ web services and eventually returns
the desired information back to the citizens. Thus, the citizens do not directly interact with the
agencies (or their web services, respectively) but only exchange information with the service
broker – creating a high level of transparency.

The exemplary architecture is depicted in Figure 12. It has partially been adapted from
[Alighieri 02] as well as from [Grimes 02].
     Please refer to [OECD]

    Government                         Government                                Government
    Agency X                           Agency Y                                  Agency Z
                   Information                         Information                                 Information
     Databases       Systems            Databases        Systems                 Databases           Systems

                                                                                                                     allocate revenues
      .NET Technologies                   .NET Technologies                       .NET Technologies

          Web Service X.1                      Web Service Y.1                        Web Service Z.1

          Web Service X.n                      Web Service Y.n                        Web Service Z.n
                        validate     publish                      pricing
                        broker       service                                       Pricing
                                                    service,     Pricing &
               PKI                                                                                       Billing &
                                                    deliver       Billing
                                       UDDI                                                             Accounting
             Authority                              results     Web Service         Pricing              System
                                                                   request /        Policies
            validate                lookup                         provide                                record
            citizen                 service                        pricing                                usage

                       E-Government Broker / Web Server (IIS) / Portal Website
                                                 service requests
                                                 pricing information

                                                   Citizen                               billing
                                                (Web Browser)

           Figure 12: Exemplary Architecture of an E-Government Service Broker

Once the citizen has communicated his needs via the broker’s portal website and the broker
has validated his identity with the help of a PKI authority, the broker looks up the required
services in an UDDI service in which all agencies have published the web services they are
able to provide. The broker then receives a WDSL message describing the data needed by the
responsible agency’s web service in order to perform the requested service. According to the
received WDSL information, the e-government broker generates an HTML form which it
provides to the citizen through the web server. After the user has completed the form, the
broker generates a respective SOAP message to invoke the web service. After validating the
broker’s identity, the web service processes the request and returns the result to the broker
encapsulated in a SOAP message. Finally, the e-government broker returns the information
contained in the SOAP message’s body to the citizen via HTTP. It is possible that in order to
fulfill a single service request from a citizen (e.g. the administration of a relocation) the
service broker has to interact with multiple agencies and web services, respectively.

8.3.4 Pricing E-Government Web Services
The .NET framework does not specify explicitly how the pricing and billing of services is to
be implemented and doesn’t provide any ready-to-use pricing modules. To solve the problem
of service pricing in the case at hand, a web service which is dedicated to pricing can be
implemented. Such a web service maintains a database where the government agencies
register the pricing policies for their services. Apart from the fixed prices for services, the
pricing policies can also contain criteria about which individuals qualify for special service
rates, e.g. low-income individuals. Discounts based on service volume as well as temporal
price discriminations may also be included. Moreover, options for premium-priced express
processing can be included in the pricing policies. Before invoking a web service from an
agency, the service broker will make use of the pricing web service to lookup the applicable
price for the requesting citizen. If multiple web services need to be invoked to solve the
citizen’s problem, the broker will look up all individual prices and present the resulting total
price to the requestor. After the citizen has approved of the price, the broker invokes the
service or services, respectively. The broker then sends a message containing service type and
price information to the accounting system which tracks the service usage for each individual
citizen. When all services have been completed, the broker requests the total amount due from
the accounting system and presents the sum to the citizen. The actual billing of the citizens is
then handled by the accounting and billing system which also allocates revenues among the
agencies according to the recorded usage. In the specific case of an e-government service
broker, it is assumed that the service broker is subsidized by the agencies and does not
generate profits itself. Thus the broker does not add a surcharge for its services. However, if
required, such a surcharge could also be administered by the accounting system.

8.4    A Model for the Exchange of Standardized Software
       Components Using Jini
In this section, we present a Jini-based architecture for the exchange of standardized software
components. We have already discussed the concept of reusable software components in
section 3.1.1 and it shall not be repeated in detail here. The basic idea is to maintain a
repository of standardized software components that provide solutions to commonly
encountered problems. These components can then be reused to build higher-level
applications. The broker architecture we describe in this section shall provide functionalities

that allow the discovery and use of software components that can be combined to construct
new applications. The focus again lies on the required pricing mechanisms and how pricing is
supported by the applied Jini technology. Similar to the two examples described previously,
the traded objects can be regarded as services that do not present their inner workings to the
user, but rather provide a defined interface that allows making use of their functionality.

                 Software Component 1
                   GIF-TIF-Converter                           Adhesive

                                             Glueserver        GIF-TIF    TIF-PNG
                 Software Component 2

              Figure 13: Basic Concept of Adhesives (adapted from [Mundt 02])

[Mundt 02] presents an architecture that enables the combination of standardized services to
provide new, integrated functionality. Mundt’s approach to service composition is to make
use of so-called Adhesives Templates. These templates contain program code which combines
existing native services to create new integrated services. The basic concept of Adhesives is
illustrated in Figure 13. Hence, Adhesives integrate the functionality of multiple software
components and present the resulting service through a single interface to the user. In the
example, the Adhesive integrates the functionalities of two independent graphics format
converters into a single service. Due to their integrative role, Adhesives can be regarded as a
type of service broker, as they facilitate the communication of users with multiple
components by providing a single interface. The following gives an overview of the
architecture presented by [Mundt 02]. Subsequently, the possibilities of integrating pricing
mechanisms into the architecture will be discussed.

                Service Provider                               Adhesive
                         Service a
                                                    4                 Adhesive x

                 Service b                                     Adhesive y

                                     Glueserver                2

                               1                               5

                               7                               6

                Client                  3

             Figure 14: Functioning of the Glueserver (adapted from [Mundt 02])

The exemplary implementation of the architecture in [Mundt 02] is based on Jini technology,
extending Jini’s Lookup Server with additional functionality to create a so-called Glueserver
which coordinates the construction of composed services. Initially, client requests for a
specific software component are handled by a Jini Lookup Server. However, if the Lookup
Server is not able to provide the client with a component that matches the client’s
requirements the request is forwarded to the Glueserver. Service providers (1) – in this case
the owners of software components – register their services with the Glueserver, which saves
and stores the describing parameters for each service. Providers of Adhesives also register
their Adhesives Templates with the Glueserver. When a client needs to make use of a specific
software component it will send a request to the Glueserver, providing a list of parameters
that specify preferences regarding the selection of the service (3). If the required software
component does not exist natively, the Glueserver queries its database to find matching
Adhesives Templates and instances these (4). The providers of the selected Adhesives
Templates then request the services specified in the templates from the Glueserver (5).
Finally, the resulting Adhesive is registered with the Glueserver (6) and the Glueserver
answers the client’s request (7). The described functioning of the Glueserver is depicted in
Figure 14.

8.4.1 Pricing Software Components in a Jini-Based Architecture
In Jini, access to resources is granted on the basis of leasing contracts. A Jini lease is a
contract between two parties, where one party is willing to grant some of its resources to the
other party. In the case at hand the grantor would be the owner of a software component that

he is willing to trade. The lease contract is valid for a limited time period only. Before
granting access to the resource, the amount of time is negotiated between the two parties,
where the grantor ultimately decides on the lease’s length. Upon expiry of the lease, the
associated resource can be freed by the lease grantor and may be used otherwise, or the lease
holder may extend the lease period [Hasselmeyer 00]. When looking at the leasing concept
with regard to the trading of software components, one has to differentiate between two cases.
In the first case, a client is in need of a software component which he will use frequently.
Hence the client will more likely be interested in obtaining permanent rights to use the
software than having to agree to a temporally limited leasing contract. In such a case, the
client will purchase the software component from the provider (if this is possible) and after
the client has obtained the component, the transaction between the provider and the client is
completed. In the second case however, the client needs to use the software component only
infrequently, maybe once a year to perform annual accounting tasks. In such a case, it is not
only cheaper to lease the software for the amount of time needed than to buy it, but it also
saves the client the time for maintaining the software and keeping it up-to-date [Hasselmeyer
00]. In the following pricing-related considerations, we assume that the latter case is

From a pricing perspective it is favorable that access to software components is handled
through a proxy using remote method invocation and that software components are not
downloaded permanently to the client’s computer. The reason for this is that remote execution
better allows controlling access to and usage of the offered software components. By being
able to strictly authorize and monitor access to software components, providers can apply
more sophisticated pricing schemes than simply charging a fixed price and allowing a client
to download the software for discretionary use. Since Jini uses the leasing concept (described
above) to grant access to resources, pricing mechanisms can also make use of this concept.
Hence prices can be set on the basis of individual lease periods. Of course various pricing
policies are applicable, e.g. discounts for long-term leases or temporal price-discrimination in
order to optimize resource usage.

The Jini technology does innately not feature any functionality to handle the pricing of
services. Additional custom features must therefore be added to equip the architecture
presented by [Mundt 02] with effective pricing mechanisms. The Jini Lookup Server does

provide methods for the registration of services. In the standard Jini implementation this is
done by indicating a proxy for the respective service. For the registration of Adhesives
Templates the interface has been extended by the Glueserver. With the Glueserver acting as a
service broker between clients and providers of software components, pricing mechanisms
shall also be implemented with its help. In a first step, service providers that register their
software components with the Glueserver also register the respective prices. These entries are
not fixed, so that service providers can update their prices when required. Providers should
also have the possibility to indicate when their resources become scarce, e.g. when their
bandwidth usage reaches a level that no longer allows the efficient provision of the software
components. Such an indication of scarcity would then require the Glueserver to raise prices
to a pre-defined level or put the Glueserver into auction-mode (i.e. collecting bids from the
requestors) for the affected component.

In the Jini architecture, clients that require a specific software component will send a list of
attributes describing their preferences to the lookup service. In the example at hand, the
Glueserver has taken over the role of lookup service. Clients can therefore also specify their
preferences regarding the price of the needed component to the Glueserver. Such preferences
may include attributes that describe the maximum price or willingness to pay a premium price
for faster delivery. Policies on how bids should be placed in case of an auction may also be
included. These attributes are then used by the Glueserver to determine the most suitable
available component.

To support pricing in the concept of Adhesives, i.e. the combination of services, the provider
of the Adhesives must be able to determine a price for its combined services. Since providers
of Adhesives make use of existing software components that are registered with the
Glueserver, the price of an Adhesive will be based on the prices of the integrated components.
Adhesive providers that are expected to purchase large volumes of a component may be given
discounts by the component’s owner, thus enabling the Adhesive provider to price the
Adhesive below the sum of the single components’ prices. Yet consumers of the Adhesive
will most likely not profit from the full extent of such a discount, as providers of Adhesives
will take their share to compensate for their efforts.

8.5    Summary: Support of Pricing Mechanisms by CORBA, .NET,
       and JINI
In the precedent sections of this work, we have discussed three examples of service broker
architectures, with each example based on another technology. As we have already mentioned
in the individual discussions of the models, none of the three technologies (CORBA, .NET,
and JINI) explicitly support pricing mechanisms. However, all three technologies are
designed in a way that they allow the additional implementation of pricing components. By
providing a Licensing and a Trading Service, CORBA does provide a foundation on which
pricing mechanisms can be implemented by adding dedicated pricing services into the
framework. JINI’s leasing concept may also be used as a basis for the implementation of
service pricing. Finally, web services may be designed in the .NET framework to provide the
required pricing and accounting functionality.

Which of the three technologies best supports the implementation of pricing mechanisms can
therefore not ultimately be determined and a recommendation for a specific framework cannot
be given. Since the realization of pricing mechanisms is possible with all three technologies,
implementers of a service broker environment will likely base their selection of the
architectural framework on other criteria.

9     A Generic Grid Architecture Supporting Multiple Pricing

In the following we shall discuss a generic grid architecture that was originally presented by
Rajkumar Buyya. The architecture was named GRACE (Grid Architecture for Computational
Economy) because it was designed specifically with the intent of implementing economic
models (including pricing mechanisms) in grid computing environments, and is therefore of
special interest to this work. After giving an overview of the architecture, we shall investigate
whether GRACE could be implemented in other existing grid environments. Furthermore, we
will analyze to what extent ASP and ISP service models can be mapped to GRACE.

9.1    Overview of the GRACE Architecture
[Buyya 02a] presents a Grid Architecture for Computational Economy (GRACE) which is a
generic architecture that can accommodate different economic models used for resource
trading and determining the service price. GRACE therefore provides services that enable
Grid Service Providers (GSPs) and user-agents to maximize their utility functions. GSPs can
use GRACE mechanisms to define their charging and access policies and have the GRACE
resource trader work according to their policies. The users are represented through Resource
Brokers that make use of GRACE services for identifying GSPs that meet their requirements.
An abstract grid architecture following the GRACE model is depicted in Figure 15.

                                                               Grid Market    Information
                                                                Services        Servers

                                                               Grid Node n
                               Grid Explorer
                                                 Secure      …
                                                            Grid Node 1
                     Job                                                     Algorithms
                                 Schedule         QoS
                    Control                                 Trade Server
      Application                 Advisor
                                                Trading                      Accounting
                              Trade Manager                  Reservation
                                                      $                       services
                        Deployment Agent                     Accounting

                                                Storage     R1    R2 R3             Rn

                     Grid Resource Broker         Grid           Grid Service Providers
     Grid User

                    Figure 15: An Abstract Grid Architecture [Buyya 02a]

The presented Resource Broker component occupies a mediating function between grid users
and GSPs using middleware services. As stated in [Buyya 02a] the Resource Broker is
responsible for resource discovery, resource selection, binding of software, data and
resources, initiating computations, adapting to the changes in grid resources virtualizing the
grid resources to present them to the user as single unified resource. The Resource Broker’s
components are described to have the following functions:

  • Job Control Agent (JCA): This component is a persistent control engine responsible for
     shepherding a job through the system. It coordinates with the Schedule Adviser for
     schedule generation, handles actual creation of jobs, maintenance of job status,
     interacting with clients and users, schedule advisor, and dispatcher.

  • Schedule Advisor (Scheduler): This component is responsible for resource discovery
     (using the Grid Explorer), resource selection and job assignment (schedule generation)
     to ensure that the user requirements are met.

  • Grid Explorer (GE): This component is responsible for resource discovery by
     interacting with the Grid Information Servers and identifying the list of authorized
     machines, and keeping track of resource status information.

   • Trade Manager (TM): This component works under the direction of resource selection
      algorithms (the Schedule Advisor) to identify resource access costs. It uses market
      directory services and GRACE negotiation services for trading with Grid Service
      Provides (i.e., their representative Trade Servers).

   • Deployment Agent (DA): This is responsible for activating task execution on the
      selected resource according to the scheduler’s instruction and periodically updates the
      status of task execution to JCA.

The broker allows users to submit their requirements by specifying a deadline and budget
(with time or cost as the optimization parameter). According to the specifications given by the
users, the broker uses scheduling algorithms to select the resources depending on their
availability, capability, and cost.

The grid middleware that is depicted in Figure 15 enables the coupling of grid users and
resources. Many of the required middleware services are readily available elements of the
Globus toolkit, such as Resource allocation and process management (GRAM), unicast and
multicast communication services (Nexus), authentication and security services (GSI),
advanced resource reservation (GARA) etc. For further descriptions of these services please
refer to [Buyya 02a) or [Globus], respectively.

Following the GRACE architecture shown in Figure 15, the Grid Service Providers are
equipped with the following components:

   • Grid Market Directory (GMD): In this directory, resource owners (service providers)
      can publish their services in order to attract customers.

   • Trade Server: This resource owner agent negotiates with Resource Brokers and sells
      access to resources. Its aim is to maximize the resource utility and profit for the GSP.
      The Trade Server consults pricing policies during negotiation and directs the accounting
      system for the recording of resource consumption and the billing of users according to
      the agreed pricing policy.

   • Pricing Policies: The prices that GSPs want to charge to users are defined in Pricing
      Policies. GSPs may make use of various pricing models described in precedent sections
      (section 7.3.1) in order to maximize their profit and resource utilization. Hence, pricing
      policies can be driven by supply and demand (such as in the commodity market model),
      can be auction-based, or can be set so that prices vary with time and from user to user.

   • Resource Accounting and Charging: These components are responsible for the tracking
      and recording of resource usage. The user is billed according to the contract between the
      Resource Broker and the Trade Server.

[Buyya 02a] defines a negotiation protocol that is used between the Trade Manager (a part of
the Resource Broker, representing the user) and the Trade Server (representing the service
provider). The basic functioning of the protocol is that the broker’s Trade Manager contacts
the Trade Server and requests a price for a specific resource. To specify the user’s
requirements so-called Deal Templates are used, which include parameters such as expected
start time, usage duration, memory and storage requirements and an initial offer. The template
is sent back and forth between the Trade Manager and the Trade Server and continually
updated until either one side decides to abort negotiations or both sides agree on a deal. It is
important to notice that both Resource Brokers and Grid Service Providers can initiate
resource trading. Resource Brokers may invite bids from a number of providers and select the
best offer, or, alternatively, service providers may invite bids in an auction and award services
to the highest bidder.

9.2    Implementing Pricing Models Using the GRACE Architecture
Possible implementations of the theoretical pricing models we have discussed in section 7.3.1
are presented comprehensively in [Buyya 02a] and shall not be fully repeated at this point.
For illustrative purposes, we shall take a look at one specific example, namely the
implementation of the commodity market model. For further implementations, please refer to
[Buyya 02a].

If pricing in a GRACE grid computing environment is implemented according to a
commodity market model, resource owning service providers will publish their services in the
Grid Market Directory. After having registered themselves as service providers, resource
owners therefore specify their pricing information so that Grid Trade Servers can use this
information for publishing the offerings in the market directory. A price specification may
consist of the following parameters:

   • consumer_id

   • peak_time_price

   • offpeak_time_price

   • discount_when_lightly_loaded (e.g. when the load is less than 50% at any time)

   • raise_price_high_demand (e.g. when the load is above 80% at any time)

   • holiday_price (price during weekends and holidays)

Resource Brokers that receive requests for resources from users will then contact the Grid
Market Directory and search for appropriate Grid Service Providers. The broker will then
contact each provider and negotiate according to the protocol we have described in section

9.3    Adaptation of the GRACE Architecture by Grid Computing
If we again look at the high level reference model for grid architecture we have discussed in
section 7.3.2 we can see that the key elements of the GRACE architecture (i.e. the Grid
Resource Broker, Grid Market Directory and the Trade Servers) can be assigned to the
collective layer. Hence, as intended by its developers, the GRACE architecture can be
implemented – at least partially – on top of existing grid infrastructures, as it does not give
any restrictions regarding implementations on lower levels.

GRACE has been designed to be compatible with existing grid middleware systems such as
Globus26, Legion27, and Condor28. Specifically, GRACE leverages the services of these
systems to add functionality that supports economic resource trading. This is done by
equipping Grid Resource Brokers with interfaces that allow the interaction with existing
middleware systems. For example, Nimrod-G is a grid broker designed according to GRACE
that performs resource management and scheduling of grid applications. In a prototype
implementation described by [Buyya 02a], Nimrod-G brokers are used to allocate resources
across five different grids that are based on different architectures.

Hence, if we are to investigate to which extent the principles of GRACE can be adapted into
existing grid environments, we must analyze the grids’ underlying architecture and put a
specific focus on the applied middleware. In the following we shall do this by discussing two
concrete examples.

9.3.1 TeraGrid
Let us first look at an example grid environment that makes use of popular grid software
solutions. TeraGrid29 is a planned computational grid aiming at deploying the world’s largest
distributed computing infrastructure. Since TeraGrid will unite various heterogeneous
computing systems, the architecture has been chosen to be as open as possible. According to
[Catlett02], TeraGrid solely defines the interfaces (protocols, schema, operational procedures)
that a site must support in order to participate in TeraGrid, without making restrictions about
the ways in which these interfaces are implemented. TeraGrid follows the approach of a
layered architecture, which can partially be mapped to the reference model presented by
[Foster 03a]. On the three layers, commonly used software components can be implemented.
For example, the Globus Monitoring and Discovery Service (MDS) may be used on the
collective layer. On the resource layer (fabric layer) the resource allocation and management
service GRAM (also part of the Globus toolkit) can be implemented. Services of the Condor
architecture are also supported. For an overview of TeraGrids service layers, please refer to
Table 7.

   For further information on the Globus toolkit, please refer to [Globus]

    Service Layer                 Functionality                   TeraGrid Implementation

    Advanced Grid Services        Super schedulers, resource      SRB, MPICH-G2, distributed
    (Application Layer)           discovery services,             accounting, etc.
                                  repositories, etc.
    Core Grid Services            TeraGrid information service,   GASS, MDS, Condor-G,
    (Collective Layer)            advanced data movement,         NWS
                                  job scheduling, monitoring
    Basic Grid Services (Fabric   Authentication and access       GSI-SSH, GRAM, Condor,
    Layer)                        Resource allocation/mgmt        GridFTP, GRIS
                                  Data access/mgmt
                                  Resource Information
                                  Local Accounting

                Table 7: TeraGrid Service Layers (adapted from [Catlett 02])

According to [Catlett02], TeraGrid does not – in its initial design – support resource
allocation according to economic models, but the primary focus is rather set on resource usage
tracking and accounting. Resources are allocated on a per-resource basis and the trade of these
allocations among users is strongly restricted. Prices of resources are fixed, i.e. there are
agreed conversion rates that allow the comparison of the relative values of resources.

We assume that the major reasons for the non-existence of economic models in TeraGrid are
the lack of practical experience in this field as well as the primary focus of development lying
on the mastering of other issues (such as global accounting). However, we believe that once
other technical problems have been mastered, TeraGrid provides an architecture that is
capable of adapting GRACE and introducing economic principles for resource allocation.
This is due to the open architecture on which TeraGrid is based, i.e. the support of open
standards and frameworks. As we have stated earlier, both TeraGrid as well as GRACE
feature support for Globus and Legion software components. If we again look at TeraGrid’s
architectural layers, we could install GRACE middleware services we have described in
section 9.1 in the Core Grid Services layer. This would imply that nodes participating in
TeraGrid would need to implement Trade Servers and other GRACE compliant services.
However, due to GRACE’s flexibility this could be achieved despite the potential
heterogeneity of the participating systems, as nodes could implement theses services in their
respective environments. Users of DataGrid on the other hand would require access to
Resource Brokers to address their requirements and be represented on the resource market.

These brokers could be implemented to the specifications of the users’ environments, as long
as these implementations would comply with the interfaces given by GRACE.

Since TeraGrid aims at uniting various computing systems that are based on highly
heterogenic environments, a strong point has been made that nodes may participate in the grid
without regard to the architectures they are based on. It is therefore important to notice that
the implementation of resource pricing mechanisms with GRACE would not require the low-
level grid implementations of the participating nodes to be altered. An adaptation of GRACE
would be handled by higher-level services, concerned primarily with the interaction between
the participating nodes, but not influencing the actual (technical) delivery of resources.

9.3.2 NorduGrid
NorduGrid30 (Nordic Testbed for Wide Area Computing) is a grid computing environment
designed to support batch processing suitable for problems in high energy physics [Eerola
03]. Geographically, NorduGrid is a distributed system with nodes in Denmark, Norway,
Sweden, and Finland. In this second example, NorduGrid shall illustrate how GRACE can be
adapted by grid environments that were built from scratch and use individual, uncommon
architectural components and software.

During the evaluation of possible architectures and frameworks to be used in NorduGrid, it
was decided not to use the standard components provided by the Globus toolkit, but rather to
extend those components to provide additional functionality, such as job brokering. Let us
quickly describe the most important components of the NorduGrid architecture (as specified
by [Eerola 03]):

      • User Interface (UI): The UI is responsible for resource discovery, brokering, grid job
         submissions, and job status querying. To achieve this, the UI communicates with the
         NorduGrid Grid Manager, Queries the NorduGrid Information System and the Replica
         Catalog. According to this architecture, NorduGrid does not make use of a central
         resource broker; instead the brokering functions are handled directly at the users’ sites.


   • Information System: This is a service within NorduGrid that provides information about
      available storage and computing resources.

   • Computing Cluster: These are the computing units in the NorduGrid architecture. Each
      cluster possesses a Grid Manager component as well as a local information service

   • Replica Catalog (RC): The RC is used for registering and locating data sources. RC
      records are updated by the GM and can be used by the UI for resource brokering.

   • Grid Manager (GM): The GM acts as a front-end for job submissions to a cluster and
      provides job management.

To attain a better understanding of the interrelations of these components, let us quickly
describe NorduGrid’s basic task flow: To find grid resources that are able to perform a
required job, a NorduGrid user describes the job with help of the Globus Resource
Description Language (RSL) and submits the job description to the UI. The UI then performs
resource brokering mechanisms by making use of the Information System as well as the data
stored in the RC. Once the UI has found an appropriate cluster, it will submit the job specified
by the user. The job is then received by the GM who ultimately submits the job for processing
and returns the results after job completion.

On the basis of the description given by [Eerola 03] we assume that NorduGrid does not
provide any mechanism for the economic pricing of resources. For the sake of our discussion,
let us assume that in a later stage NorduGrid shall adapt the concepts of GRACE to introduce
market mechanisms for resource allocation.

In the GRACE architecture, individual users are represented through Grid Resource Brokers
to acquire the required resources. In NorduGrid, users have access to User Interfaces that
provide brokering functions. However, to adapt the GRACE principles, UIs would need to be
expanded to support pricing and bidding preferences in their job specifications. In order to do
so, the used specification language, RSL, would need to be extended as well. Other
functionalities that the GRACE architecture assigns to the Grid Resource Broker – such as
Job Control Agent, Schedule Advisor, and the Grid Explorer – are already inherent to the UI
or are performed in conjunction with the Replica Catalog or the Information Service.

On the side of the cluster nodes, Grid Managers would take over the role of the Trade Servers
depicted by the GRACE architecture. GMs would therefore need to define pricing policies for
the resources for which they are responsible. These pricing policies could be stored in the
Replication Catalog and provided by the Information Service, which provide the functionality
that is equivalent to the Grid Market Directory in the GRACE architecture. A further
requirement would be the introduction of a common bargaining protocol, to allow price
finding between UIs (Trade Managers) and GMs (Trade Servers). To enable a resource
market, the current monitoring mechanisms would need to be extended to provide sufficient
accounting and billing functionality.

Furthermore, we shall again notice that the adaptation of market mechanisms does not
influence the lower level hardware implementation of the involved computing clusters, as the
Grid Managers featured by NorduGrid’s original architecture provide the required abstraction

9.4      Mapping ISP and ASP Service Models to the GRACE
Let us now investigate to which extent the service models of Internet Service Providers (ISPs)
and Application Service Providers (ASPs) we have discussed in the seventh chapter of this
work can be mapped to the GRACE architecture. To do this, we must first examine whether
the services of ASPs and ISPs can be provided through grid infrastructures.

Essentially, ASPs provide access to software solutions via computer networks. From this
point of view, one could argue that this essentially the same as the application services that
are provided in grid computing. However, the difference lies in the type of software that is
provided. As we have mentioned in the second chapter of this work, ASPs often distribute
complex software that is highly customized for individual customers. Hence, the level
commoditization is very low. Moreover, in the “traditional” ASP model, contracts between
providers and customers tend to be middle- or long-term, due to the high investment required
by the customization. If we look at grid computing, things tend to be contrary. In grid
computing environments, resources are highly commoditized so that they can be exchanged

easily. Moreover, relations between providers and users are ad-hoc style, short-term contracts,
terminated as soon as the resources required by the user have been delivered. Following our
attention to map the ASP service model to GRACE, we must assume that ASPs also provide
applications that can be distributed via a grid infrastructure. Yet, if we assume that ASP
software is completely commoditized, a further distinction between an ASP and regular Grid
Service Providers would no longer be possible. Hence, let us assume that ASPs provide some
type of software that is compatible with provision via a grid, yet it is not commoditized to the
point where it is provider neutral. If we look at this scenario with regard to GRACE, we will
find an environment that is very similar to the service broker architecture we have discussed
earlier in this work. Hence, ASPs would register their services in a Global Market Directory
(as foreseen by GRACE) via a Trade Server they have installed at their site. Users on the
other hand would make use of Service Brokers to represent their requirements and would
eventually be matched with a convenient provider. We assume that in comparison to the
“traditional” ASP service model, ASPs would have to adapt to stronger competition, due to
the presence of service brokers. Moreover, if the provided software is partially commoditized
and all market participants are connected via a grid architecture (which would be the case in a
GRACE environment), we could think of a scenario where users would access certain
components of provided software from one provider and other components from others.
Whether this would be possible would depend on the granularity of the software that is being
provided. With regard to pricing, the services provided by ASPs could be priced according to
the ASP pricing schemes we have presented in section 7.2.4. ASPs would specify their
pricing policies and publish them in the Grid Market Directory, as foreseen by the GRACE
architecture. Hence with the discussed restrictions regarding service commoditization, we can
apply ASP service models in a GRACE environment.

If we now look at the service model of ISPs and its compatibility with grid computing
architectures, we have the special case that ISPs actually provide a service that is a
prerequisite for any grid to function. ISPs provide the basic connectivity that is needed in
order to connect to a grid. It is therefore questionable to which extent such services can be
traded on a grid market or are rather agreed upon through traditional channels. Yet other
services, such as special connectivity (high bandwidth, guaranteed quality etc.) or hosting
(provision of storage) services are well commoditized goods and can therefore be traded and
deployed through grid computing environments. As these services closely resemble those of

Grid Service Providers, a further discussion of how these services can be mapped into a
GRACE environment becomes redundant.

9.5    Conclusion: Realizability of a Global Economic Market for Grid
A prototype based on the GRACE architecture has been implemented and tested successfully,
as described in [Buyya 01a]. However, the prototype featured several limitations, for example
prices were not truly dynamic but were fixed after initial scheduling decisions had been made.
Yet, in order to discuss the realizability of a global computational grid in which resources are
allocated through market mechanisms following economic principles, we have to take into
consideration higher-level issues. Moreover, we need to concentrate on the problems that
have to be solved specifically with regard to the realization of a grid that follows economic
principles, i.e. issues of organizational, economic and political nature. Regarding purely
technical issues, the interested reader may find a discussion of a variety of general technical
requirements that any grid architecture must fulfill in [Grimshaw 03].

A first prerequisite that needs to be satisfied in order to realize a global market for grid
resources is the availability of grid resources in the form of commoditized goods, i.e.
resources need to be available with uniform characteristics and supplier neutral. Moreover,
these commodities must be provided with guaranteed quality of service properties, because –
according to [Kenyon 03] – uncharacterized, unguaranteed resources have no value.

If we assume that commodity goods are available for trading and the above criteria can
therefore be fulfilled, questions about the organization of the aspired market arise. Would the
market be able to regulate itself? Or would it require some regulating authority? And if the
latter is the case, who or what would form that authority? Where would it reside? Which
competencies would it have? Would it, for example, also be empowered to decide on
technical questions, such as the setting of standards?

An even more fundamental question is whether the grid market place would need to be truly
global and centralized or if not perhaps a number of distributed – yet interconnected –

marketplaces would be more convenient. Moreover, one must ask whether there is a critical
size for the aspired market, i.e. how many participants would be required in order for price
mechanisms to function properly or whether there would be a maximum of participants the
market could cope with.

Yet another interesting aspect is introduced in [Wolski 03]. Regarding the realization of a
global grid economy, [Wolski 03] argues that a key requirement for grid economies in general
is that grid systems will need to be partially or entirely reset, e.g. to perform software or
hardware maintenance or for security concerns. Yet, according to [Wolski 03] the economics
research community has only little knowledge about starting an economy from scratch,
making the influences of such resets on economic markets hard to predict. Hence we can
conclude that such uncertainties yield risks for market stability and may hinder the acceptance
of global grid markets.

If we are to construct a grid architecture that supports a variety of pricing models, we must
ask in whose interest these pricing models are, as this has an impact on the acceptance of the
models and finally results in whether the market is used or not. For example, service
providers may prefer the usage of auction-based pricing mechanisms, as they tend to achieve
high prices. Users might not be willing to go through a bidding process for every resource
request and prefer fixed prices. Hence, how large a variety of available pricing mechanisms
an acceptable grid architecture needs to support would have to be investigated. This leads us
to the general question about the degree of complexity a pricing mechanism can possess
without losing its practicability. While a complex pricing mechanism might be efficient from
the economic point of view in the way that it finds a price that is Pareto efficient, the resulting
overhead caused by negotiation and administration might render the pricing model useless.

Another block of question arises when we consider how a global resource trading network,
such as the aspired grid economy, would relate to the real world economy. We can assume
that the grid market would require some kind of virtual currency. How would such a currency
relate to its various real-world counterparts? Moreover, how could the virtual settlement of
contracts that may involve numerous trading partners be reflected in the real world?

Furthermore, what effects would real world events, such as natural disasters have on the
virtual market? How could these effects be managed?

As we can see from the questions above, there are numerous uncertainties and issues to be
resolved. Some of them yield very elementary problems whose complete discussion would
exceed the scope of this work. However, as we cannot answer these questions here, they shall
illustrate the complexity that has to be mastered. Moreover, what emanates is the notion that
the creation of a global grid market involves the solution of technical, economic,
organizational as well as political questions. It is therefore crucial that a coordination of
various developments that are made in different areas is achieved, so that the progress made
can propel the advances towards the vision of a global grid economy.

10 Conclusion

10.1 Summary
We have noticed that increased service orientation can currently be observed in two separate
domains and that there are parallels between the two. On the one hand, service orientation is
an important topic for IT organizations that consider themselves service providers for
corporate internal and external customers and want to provide well-defined IT services that
are easily manageable to users. On the other hand, service orientation evolved into a paradigm
for software development and engineering, in which applications and resources within a
network are regarded as services. What these two developments have in common is that they
both attempt to reduce the complexity of intricate systems by providing end users with high
system transparency, in order to facilitate their access to and use of the available services.
Hence users are not bothered with details about how the results of the desired service are
produced, but are rather provided with well-defined functionality that corresponds to their

In this thesis we have investigated how and in what way pricing of such services can be
achieved. In a comprehensive approach, we have investigated economic, managerial,
organizational and technological issues, attempting to identify as many pricing-relevant
factors as possible. Amongst others, we have:

  • discussed various types of available services and described the characteristics of service
     providers that offer them.

  • reviewed the economic principles of price and identified the specialties of pricing IT-
     based services.

  • looked at the various participants of service markets, illustrated their interests and
     analyzed their influence on price.

  • presented technologies and organizational concepts that are relevant to the development
     of service-oriented IT environments.

  • presented a number of theoretical pricing models for IT-based services and looked at the
     pricing models and schemes applied by service providers in practice.

  • identified current real-world limitations for theoretical pricing models and defined the
     resulting requirements on technology and administration.

  • discussed the possibilities of pricing services in markets that are organized through
     service brokers and developed concrete application scenarios that make use of different

  • analyzed the differences between pricing for various types of resources as well as the
     dissimilarities of pricing in scientific and commercial environments in grid computing.

  • presented a grid computing architecture that allows resource allocation according to
     economic principles by supporting different pricing models, and mapped multiple
     existing grid architectures and other models of service provisioning to the architecture.

10.2 Key Findings
In order to apply the concept of service-oriented IT successfully in practice, feasible pricing
mechanisms are required. A substantial amount of academic research has been devoted to the
development of pricing models for Internet services, which is reflected in the large variety of
models available. However, we have observed a latent discrepancy between theoretical
pricing models and how pricing is performed in practice. We believe that the reason for this
lies in the complexity of the theoretical models that render these infeasible in practice. The
problem yields from the fact that theoretical models often attempt to model reality by
integrating parameters that cannot be determined precisely in the real world. In general, a
trade-off between accuracy and usability seems apparent for the development of any pricing

A corporation’s decision to organize its IT environment according to the concepts of service
orientation will have extensive consequences and as well as an impact on corporate culture. IT
departments will be considered internal service providers that treat business functions in a
customer-oriented manner, as they support the business by providing well-defined and
manageable services. Tools such as service level agreements and service management
frameworks support the efficient communication between IT organizations and business and
enable the former to show the value they provide to the corporation and demonstrate that IT
does matter.

We have noticed a trend in software development away from monolithic applications towards
software components that offer their functionality in the form of services. The enabling
standards and technologies are referred to as web services. From an economic point of view,
the trend towards standardized components reflects a commoditization of the products. This
allows the introduction of new market models, as we have seen when we discussed the trading
of standardized software components with the help of service broker architectures or in grid
computing environments. Software producers could therefore the longer the more be put in
the role of service providers, as they offer software functionality via electronic networks. The
services of the various providers are then combined by service brokers to provide customized
solutions to the consumer. With the technological development being strongly driven forward,
we believe that the commercial application of web services will soon flourish, i.e. that
individual providers will be able to offer web services and effectively charge prices for their
efforts. However, regarding the realization of a global market in which web services of
multiple providers are traded freely we are more pessimistic, as apart from the lack of feasible
economic models many organizational issues need to be resolved.

The concept of service brokers has accompanied us throughout this work. In their
intermediating function of matching suppliers with consumers, they are the central element in
service markets. As we have demonstrated by developing concrete application scenarios,
different technological frameworks (CORBA, JINI, and .NET) support the implementation of
pricing mechanisms in environments served by service brokers. However, concrete services
dedicated to provide pricing functionalities for other services need to be developed.

There are existing models for implementing economic concepts, including pricing
mechanisms, into grid computing environments. However, the elaborated frameworks are
very high-level and the present prototypical implementations have substantial functional
limitations. Despite the fact that there is extensive research being performed in the field of
grid computing and in the area of grid economics in particular, it is doubtful whether the
vision of a global grid service economy can be realized in the near future. Apart from the
technological developments, we believe that other issues must first be resolved before such a
market can become reality. First of all, we must ask ourselves who would be interested in the
establishment of a market for computational resources powered by grid technology, as it is
necessary that a potent organization drives the creation of the market. This also requires that

potentially large consumers of the resources that are to be the traded – e.g. larger
corporations, and academic institutions – can be convinced that participating in a grid
economy is (economically) more convenient than acquiring resources in traditional ways.
Hence, such institutions must be presented with a profitable business case, which – to our
knowledge – has not been established so far. Another strategy would be to target the
consumer market, as a global grid market in which individuals provide and consume
resources in a peer-to-peer fashion would certainly be subject to network effects. However,
for the network effects to become effective and thus enable the conquest of the consumer
market, one is in desperate need of a killer-application. The free sharing of digital content in
peer-to-peer networks has been a phenomenal success among consumers and we could think
of a scenario in which individuals would offer their unused computational resources on a grid
market. However, commercial applications would require not only effective pricing
mechanisms, but also practicable billing solutions – both of which are yet to be developed.
Moreover, as soon as services are applied commercially, the expectations and requirements on
service quality increase, which is a serious issue with regard to the best-effort quality current
Internet services are based on. For these reasons we believe that the realization of a global
market for computational resources using grid computing technology will – for now – remain
not more but a fascinating utopia.

10.3 Future Research

Support of Pricing Mechanisms by Service-Oriented Technologies

In this thesis, we have discussed pricing models that are to be applied in technologies which
are currently still strongly under development. Hence, further technological research in the
fields of web services and grid computing is certainly required. However, from the point of
view of this thesis, a stronger focus on the economic perspective during the development of
these technologies is desirable. Specifically, further research is required if the discussed
technologies are to be commercially applied in a broader sense. For example, the existing
service broker frameworks and grid toolkits need to be expanded to explicitly support a
variety of pricing mechanisms. This might collide with the general interest to keep these
service-oriented frameworks as open and universal as possible. Thus, to which extent
protocols and service components can be standardized to facilitate the pricing process – and
ultimately enable automated exchange of services among heterogeneous systems – without

substantially limiting the areas in which they can be applied requires further investigation. A
starting point for such research could be a detailed analysis of the requirements pricing
models and mechanisms have on technology. Moreover, it would have to be determined
which components and interfaces of the systems participating in a market require
standardization in order to enable the economic exchange of services. This could be done with
regard to the high-level architectures we have presented in this work. The economic
requirements on standards would then have to be judged from a technological point of view,
assessing their practical realizability. Technologically feasible requirements would then be
mapped against the existing standards in order to identify the gaps in which further
standardization is required. The further development would then need to involve close
cooperation of economic and technological experts in order to ensure both economic
adequacy as well as technological realizability of the evolving standards.

Allowed Complexity of Practical Pricing Models

As we have indicated in the key findings above, a discrepancy between the theoretical pricing
models and the pricing mechanisms used in practice can be examined. A key question in the
further development of pricing models would therefore be how complex a pricing model is
allowed to be so that it is still feasible in practice. Consequently, future research should
examine which factors need to be considered and integrated into a pricing model in order to
effectively model market behavior, under the constraint that all required parameters can be
efficiently determined in reality and that transaction costs (e.g. the administrative overhead)
are minimized. The results of such research could consist of a pricing toolkit, which would
contain a number of pricing model components that could be combined to construct pricing
models that feature the most suitable cost-value ratio for a given market environment. Hence,
the toolkit could be used to construct pricing models that could be applied to determine a
price quickly and at low cost (i.e. low administrative overhead, low cost to determine input
parameters, etc.); whereas at the same time it would be possible to add additional components
to more accurately model the market, but at increased cost. To realize such a toolkit, basic
pricing components (e.g. a component that could model demand in dependence of changes in
price) would need to be identified, and the cost of implementation would need to be
determined for each component. Maybe there would be components that need to be included
into every pricing model, whereas the implementation of other components would be so
costly that an application would only be reasonable for environments in which price remains

fixed for a relatively long time period. Of course, profound investigation whether the accuracy
of composed pricing models can be sufficient would also be required.

Feasible Market-Economy Pricing and Management Models for Internal Service

In the third chapter of this thesis we have observed that in practice the implementation of
effective IT service charging has proven to be very difficult. While there are models and
service management frameworks that describe how fair and transparent charging and
accounting of services can be managed, the generated administrative overhead often leads to
the selection of much simpler methods, such as per-seat charging. Such simple methods may
well be highly practicable and inexpensive in use; however, they do not fulfill the criteria of
being truly service-oriented, as they lack the ability to charge users explicitly for their used
services instead of confronting them with a general lump-sum payment. Moreover, price
calculations are performed cost-based, i.e. actual supply and demand for a particular service is
not considered in the pricing process. Further research in this are should therefore investigate
the reasons for the lack of application of economic service management frameworks in
practice. Despite the impact on administration, the effects on corporate policy and culture
should also be analyzed in order to identify further causes of hindrance. Once these factors
have been analyzed it should be examined to which extent existing service management and
pricing frameworks could be altered in order to achieve higher acceptance. To which extent
market based pricing models could be introduced should also be investigated. Hence, models
that allow internal pricing and charging of services according to actual supply and demand
should be developed and feasibility of use within corporate organizations should be evaluated.
Furthermore, we have noted in section 5.6.1 that internal service charging is often dependent
on political issues. From a management point of view, it would therefore be interesting to
examine whether there are feasible management models for the internal pricing of services
under the constraints of corporate policy.

Corporate IT Service Broker

In future research, it would further be interesting to investigate whether the service broker
concept could also be applied to corporate IT. A corporate IT service broker would then be
used to maintain the corporate IT service portfolio. Corporate business functions would
provide the service broker with their IT-related requirements and the broker would then
choose from the available internal and external sources to provide the required functionalities.
Hence the broker would be the business’ point of contact regarding IT services and would at
the same time represent the business’ interests towards internal and external service providers.
Ultimately, it would be the task of the service broker to optimize the corporation’s IT service
portfolio by acquiring and managing the required services at the lowest possible cost. The
optimization of the IT portfolio would resemble an on-going management process performed
by the service broker, as it would constantly evaluate the current service portfolio with regard
to the business’ requirements and the available service offerings. The roles and tasks of the
service broker are depicted in Figure 16.

                                       Service Broker
                                       •analyze business’ IT requirements
                        Requirements   •aggregate requirements                                Internal &
          Corporate                    •analyze service offerings
                                       •match requirements with offerings        Bargaining    External
          Business                     •provide business with customized                       Service
          Functions                    services
                           Services                                                           Providers
                                       •point of contact regarding IT services
                                       for business
                                       •representative of business towards
                                       service providers
                                       •manager / optimizer of IT service


                                            IT Service Portfolio

                      Figure 16: Draft of a Corporate IT Service Broker

Apart from the technological implementation, it would be necessary to examine how such a
broker could be set up organizationally, i.e. one would have to investigate questions such as

whether a broker could be implemented centrally or whether a decentralized setup would be
more appropriate. Which instance would ultimately have the competency to make sourcing
decisions would also need consideration. Moreover, the effects on corporate culture would
need to be analyzed, as corporate-internal IT departments would be forced to compete against
external service providers. Hence, such a development would somewhat resemble the
introduction of internal market mechanisms. However, while it has been commonly agreed
upon the fact that corporations serve to reduce transaction costs by replacing market
mechanisms with hierarchy, corporations that depend on hierarchy too strongly run the risk of
becoming inefficient. In general, a study on the extent to which market mechanisms can be
introduced inside a corporation to achieve efficient service provision should therefore also be
part of future research.

11 Appendix

11.1 Task Description / Aufgabenstellung

Diplomarbeit für Herrn cand. inform. Michael Werder

Service Pricing for ISPs, ASPs, and in Grid-Computing

1. Diskussion der Bedeutung des „Dienst“-Konzepts in der IT-Praxis (1/4)
   • terminologisch – hier sollte zuerst ein kurzer, historischer Überblick über
      unterschiedliche     Definitionen    in    der   Literatur   und    unterschiedliche
      Bedeutungszuweisungen in der Praxis gegeben werden und daraus sollten dann klare
      Definitionen für IT-Dienste im Allgemeinen und Dienste von ISPs, ASPs und im
      Gridcomputing im Besonderen abgeleitet werden, welche auch durch Praxisbeispiele
      erläutert werden sollten
   • aus technischer Perspektive – hier sollte unter anderem die Bedeutung für den IT-
      Architekturentwurf und der Zusammenhang mit SW-Komponenten erläutert werden
      und es sollte auch die Frage diskutiert werden, ob die Bedeutung von Client/Server
      Technologie angesichts von Peer-to-Peer Technologie und mobilen Agenten
      zurückgehen wird
   • aus Managementperspektive – hier sollte unter anderem das Qualitäts- und
      Kostenmanagement und das Reporting, Controlling und Accounting diskutiert werden
   • aus Kommunikationsperspektive – hier sollte unter anderem die Bedeutung des
      Konzepts für interdisziplinäre Kommunikation und Kooperation innerhalb eines
      Unternehmens zwischen IT-Experten, deren Kunden und dem der Geschäftsleitung
      diskutiert werden und es sollte auf Sprachregelungen wie „unsere IT-Abteilung ist ein
      Profit-Center, keine Kostenstelle“ eingegangen werden
   • aus ökonomischer Perspektive – hier sollte unter anderem die Bedeutung des
      Konzepts für das IT-Outsourcing und den Handel von IT-Dienstleistungen diskutiert
      werden, wobei auch auf die Zukunfsperspektive „standardisierte Services sind
      Commodities, die von Verzeichnisdiensten kombiniert und vermittelt werden“
      eingegangen werden soll
2. Verfassen eines theoretischen Überblicks über generelle Preisbildungsmöglichkeiten für
   IT-basierte Dienste oder Diensterbringung (1/3)
   • Überblick über unterschiedliche Ziele und Grundkonzepte für die Preisbildung,
      Einflussfaktoren für den Preis (z.B. Markt, Kosten, Name, Risiko, Art des Vertrags,
      etc.), Modelle und Verfahren zur Preisbildung
   • Framework zur Darstellung der unterschiedlichen Aspekte der Preisbildung,
      Zusammenhänge zwischen den einzelnen Aspekten und Besonderheit der Preisbildung
      für IT-basierte Dienste
   • Klassifikation der unterschiedlichen Verfahren und Einordnung in den Framework
   • Besonderheiten der Preisbildung für den Dienst eines dynamischen virtuellen
      Unternehmens, für Reservationen, für Leistungsgarantien und Risikoübernahme und
      für mittel- und langfristige Dienstverträge
   • Unterschiedliche Perspektiven aus Anbieter-, Vermittler-, 3rd Party und Nutzersicht

   •   Einschränkungen für die Anwendbarkeit von Modellen und Verfahren und
       Anforderungen an die IT-Infrastruktur und die Administration – hierbei sollte nicht
       nur Probleme bei der eigentlichen Preisbildung berücksichtigt werden (z.B. inexistente
       Arbeitslastinformation beim Outsourcing), sondern auch Probleme von Verrechnung
       und Administration
   •   Nebenwirkungen der verschiedenen Verfahren auf den Markt, das Produzenten und
       das Konsumentenverhalten
   •   Kriterien für den Erfolg eines Verfahrens und, daraus abgeleitet, Kriterien für die
       Selektion eines konkreten Verfahrens
   •   Bedeutung von Brokern und von Standards für einen Dienstmarkt
3. Diskussion der Besonderheiten der Preisbildung für ISPs, ASPs und im Gridcomputing
   • Literaturüberblick       über      Preisbildungsmodelle,   Systemarchitekturen      und
       Organisationsmodelle, die diese realisieren und Überblick über in der Praxis
       eingesetzten Verfahren und die damit gemachten Erfahrungen
   • Preisbildungssmodelle für Servicebroker, beispielhafte Diskussion der praktischen
       Realisierbarkeit anhand einer CORBA-Architektur für Leistungsaustausch in einem
       Unternehmen, einem E-Government Service-Broker und einer Jini-Architektur für den
       SW-Komponentenhandel mit kombinierten Standardkomponenten – die Beispiele
       sollten nicht nur in Bezug auf die Technologie sondern auch in Bezug auf die
       Darstellung des Geschäftsfelds möglichst konkret sein – und Diskussion des
       Einflusses der Existenz von maximaler Vernetzung und Broker-Diensten oder Grid-
       Infrastrukturen auf die Preisbildung
   • Preisgestaltung bei SLA-basierter Diensterbringung und beispielhafte Diskussion der
       praktischen Realisierbarkeit anhand von Praxisbeispielen aus der Wirtschaft – hier
       sollte auch auf Erfahrungen und Nebeneffekte eingegangen werden – nebst einer
       Diskussion des Einflusses der Existenz von Service und SLA-Standards auf die
   • Unterschiede der Preisbildungen im Grid-Computing in den fünf Bereichen Daten,
       Kommunikation, Rechnen, Werkzeuge, Wissen und zusammengesetzte Dienste
   • Unterschiede der Preisbildung in universitären Forschungsgrids und in kommerziellen
       Grids, sowie Möglichkeiten und Probleme der Koexistenz von akademischen und
       kommerziellen Preisbildungsmodellen
   • Entwurf einer generischen Grid-Architektur, die eine grosse Palette von
       Preisbildungsmodellen technisch unterstützt, Abbildung existierender und geplanter
       Grids in diese Architektur hinein, Abbildung von ISP und ASP Dienstmodellen in
       diese Architektur hinein und Diskussion der Realisierbarkeit einer solchen Architektur
       in der Praxis
   • Unterschiede der Preisbildung bei ISPs, internen und externen ASPs und in diversen
       (anderen) Formen des Gridcomputing

4. Zukunftsprognosen (fakultativ)
   • Zukunftschancen der verschiedenen Preisbildungsmodelle und –verfahren, sich auf
      dem Markt durchzusetzen und zu erwartende Auswirkungen auf die Entwicklung von

                                                          Dr. Reinhard Riedl

12 References

[Alighieri 02]   Alighieri, Giulio, “Global E-Government Web-Services”,
                 Master’s Thesis, University of Zurich, 2002

[Altmann 00]     Altmann, J.; Rupp, B.; Varaiya, P., ''Effects of Pricing on Internet User
                 Behavior,'', NetNomics, Vol.3 No.1, 2000

[Altmann 01]     Altmann, J.; Chu, K., “How to Charge for Network Services – Flat-Rate
                 or Usage-Based?“, University of California at Berkeley, 2001

[Anderson 03]    Anderson, Spencer, “Business + IT = Resilience”, Contingency Planning
                 & Management, September / October Issue, 2003

[Arnold 89]      Arnold, D.R.; Hoffman, K.D.; McCormick, J., “Service pricing: a
                 differentiation premium approach”,
                 Journal of Services Marketing, Vol. 3, 1989

[ATIS]           American National Standard for Telecommunications –
                 Telecom Glossary, 2000

[ATKearney]      A.T. Kerney & The Stencil Group, “The Emerging Web Services
                 Market”, Report, 2002

[Aviram 03]      Aviram, Amitai, “Regulation by Networks”,
                 Working Paper, University of Chicago, March 2003

[Bailey 97]      Bailey, J.P.; Bakos, Y., “An Exploratory Study of the Emerging Role of
                 Electronic Intermediaries”,
                 International Journal of Electronic Commerce, Vol. 1 No. 3, 1997

[Barmouta 03]    Barmouta, A.; Buyya, R., “GridBank: A Grid Accounting Services
                 Architecture (GASA) for Distributed Systems Sharing and Integration”,
                 17th Annual International Parallel & Distributed Processing Symposium
                 (IPDPS 2003) Workshop on Internet Computing and E-Commerce, 2003

[Benjamin 02]    Benjamin, Leon, “Web Services Business Models”, DestiCorp Ltd, 2002

[Berman 03]      Berman. F.; Fox, G.C.; Hey, A.J.G., “Grid Computing – Making the
                 Global Infrastructure a Reality”, John Wiley & Sons Ltd, London, 2003

[Bernard 03]     Bernard, Allen, “Corporate IT as ASP”, IT Management: CIO Strategies,

[Berstis 02]     Berstis, Viktors, “Fundamentals of Grid Computing”, IBM Redbooks
                 Paper, IBM Corporation, 2002

[Bichler 02]     Bichler, M., “Applications of flexible pricing in business-to-business
                 electronic commerce”, IBM Systems Journal, Vol. 41 No. 2, 2002

[Black 73]       Black, F.; Scholes, M.; “The Pricing of Options and Corporate
                 Liabilities”, Journal of Political Economy, Vol. 81, 1973

[Bloomberg 03]   Bloomberg, Jason, “The role of the service-oriented architect”, The
                 Rational Edge, May 2003, Rational Software, 2003

[Bolton 02]      Bolton, G.; Katok, E.; Ockenfels, A., “Bridging the Trust Gap in
                 Electronic Markets - A Strategic Framework for Empirical Study”, Max
                 Planck Institute for Research into Economic Systems, Germany, 2002

[Boulding 68]    Boulding, K., “Knowledge as a commodity”, Beyond Economics: Essays
                 on Society, Religion and Ethics, University of Michigan Press, 1968

[Brennan 02]     Brennan, Jack, “Show Me the Value”, CIO Magazine, May 1 Issue, 2002

[Buyya 01]       Buyya, R.; Stockinger, H.; Giddy, J.; Abramson, D.,”Economic Models
                 for Management of Resources in Peer-to-Peer and Grid Computing”,
                 Monash University, Australia / CERN, Switzerland, 2001

[Buyya 01a]      Buyya, R.; Abramson, D.; Giddy, J., “An Economy Grid Architecture for
                 Service-Oriented Grid Computing”, Monash University, Australia, 2001

[Buyya 02]       Buyya, R.; Abramson, D.; Giddy, J.; Stockinger, H., “Economic Models
                 for Resource Management and Scheduling in Grid Computing”, in
                 Concurrency and Computation: Practice and Experience, John Wiley &
                 Sons, Ltd., 2002

[Buyya 02a]      Buyya, Rajkumar, “Economic-based Distributed Resource Management
                 and Scheduling for Grid Computing”, Dissertation, Monash University,
                 Melbourne, Australia, 2002

[Cannady 01]     Cannady, S.; Stockton T.H., “Easing the PAIN - How PKI can reduce the
                 risks associated with e-business transactions”,
                 IBM Developer Works, 2001

[Catlett 02]    Catlett, Charlie, “The TeraGrid: A Primer”, 2002

[Cheng 02]      Cheng, H.K.; Koehler, G.J., “Optimal Pricing Policies of Web-Enabled
                Application Services”, University of Florida, 2002

[Chess 94]      Chess, D.; Harrison C., Kershenbaum A., “Mobile Agents: Are They a
                Good Idea?”, IBM Research Report, 1994

[Clark 97]      Clark, David D., “Internet Cost Allocation and Pricing”, in Internet
                Economics, ed. McKnight and Bailey, MIT Press, 1997

[Condor]        Condor – High Throughput Computing

[CORBA]         Common Object Resource Broker Architecture

[Dearden 78]    Dearden, J., “Cost accounting comes to service industries”,
                Harvard Business Review, Vol. 56, 1978

[De Laat 01]    De Laat, Paul B., “Emerging roles for third parties in cyberspace“,
                Ethics and Information Technology, Vol.3 No. 4, 2001

[De Roure 03]   De Roure, D.; Baker, M.A.; Jennings, N.R.; Shadbolt, N.R., “The
                evolution of the Grid”, Grid Computing – Making the Global
                Infrastructure a Reality, John Wiley & Sons Ltd, London, 2003

[Desiraju 99]   Desiraju, R.; Shugan, S.M., “Strategic Service Pricing and Yield
                Management”, Journal of Marketing, Vol. 63, No. 1, 1999

[Dewire 00]     Dewire, D.T., “Application service providers”,
                Information Systems Management, Vol. 17 No. 4, 2000

[Eerola 03]     Eerola, P.; Konya B.; Smirnova O.; Ekelöf, T.; Ellert, M.; Hansen, J.R.;
                Nielsen, J.L.; Wäänänen, A.; Konstantinov, A.; Ould-Saada, F., “The
                NorduGrid architecture and tools”, 2003

[Ferris 03]     Ferris, C.; Farrell, J., “What are Web Services?”,
                Communications of the ACM, Vol. 46 No. 6, 2003

[Fishburn 97]   Fishburn, P.C.; Odlyzko A.M.; Siders, R.C., “Fixed fee versus unit
                pricing for information goods: competition, equilibria, and price wars”,
                First Monday, Vol. 2 No. 7, 1997

[Franklin 03]   Franklin, Curtis Jr., “Grid-dy determination”,
                Network World, Vol. 20 No. 1, 2003

[Focacci 03]     Focacci, L..; Mockler, R.J.; Gartenfeld M.E.; Dologite, D.G., “How to
                 choose an ASP: selection Guidelines”;
                 Information Management & Computer Security, Vol. 11 No. 2, 2003

[Foster 03]      Foster, I.; Kesselman, C.; Nick, J.M.; Tuecke, S., “The physiology of the
                 Grid“, in Grid Computing – Making the Global Infrastructure a Reality,
                 John Wiley & Sons Ltd, London, 2003

[Foster 03a]     Foster, I.; Kesselman, C.; Tuecke, S., “The anatomy of the Grid“, in Grid
                 Computing – Making the Global Infrastructure a Reality, John Wiley &
                 Sons Ltd, London, 2003

[Fou 01]         Fou, John, “Web services and mobile intelligent agents – combining
                 intelligence with mobility”,, 2001

[Froomkin 96]    Froomkin, Michael, “The Essential Role of Trusted Third Parties in
                 Electronic Commerce”, University of Miami School of Law, 1996

[Furht 00]       Furht, B.; Phoenix, C.; Yin, J.; Aganovic, Z., “An Innovative Internet
                 Architecture for Application Service Providers”, Proceedings of the 33rd
                 IEEE International Hawaii Conference on System Sciences, 2000

[Game Theory] – Resource for Educators and Students of Game Theory

[Geppert 97]     Geppert, A.; Kradolfer, M.; Tombros, D., “Market-Based Workflow
                 Management“, Technical Report 97.15, Department of Computer
                 Science, University of Zurich, 1997

[Glidewell 00]   Glidewell, Richard A., “Business lessons from online porn”,
                 Upside, Vol. 12 No. 4, 2000

[Globus]         The Globus Alliance (home of the Globus Toolkit)

[Grimes 02]      Grimes, Fergal, “Microsoft .NET for Programmers”,
                 Manning Publications Co., 2002

[Grimshaw 03]    Grimshaw, A.; Natrajan, A.; Humphrey, M.A.; Lewis, M.J.; Nguyen-
                 Tong, A.; Karpovich, J.F.; Morgan, M.M.; Ferrari, A.J., “From Legion to
                 Avaki: the persistence of vision”, in Grid Computing – Making the
                 Global Infrastructure a Reality, John Wiley & Sons Ltd, London, 2003

[Groth 95]       Groth, John, “Important factors in the sale and pricing of services”,
                 Management Decision, Vol. 33 No. 7, 1995

[Guiltinan 87]   Guiltinan, J.P., “The price bundling of services: a normative framework”,
                 Journal of Services Marketing, Vol. 51, 1987

[Hapgood 03]       Hapgood, Fred, “Plug and Pay”, CIO Magazine, April 15 Issue, 2003

[Hasselmeyer 00]   Hasselmeyer, P.; Schumacher, M.; Voss, M., “Pay as You Go –
                   Associating Costs with Jini Leases“, Darmstadt University of
                   Technology, Germany, 2002

[Hoffman 89]       Hoffman, K.D.; Arnold, D.R., “Professional service pricing: an extended
                   cost-oriented approach”,
                   Journal of Professional Services Marketing, Vol. 5 No. 1, 1989

[Hoffmann 03]      Hoffmann, K. Douglas, “Marketing + MIS = E-Service”,
                   Communications of the ACM, Vol. 46 No. 6, 2003

[Hofmann 99]       Hofmann, H.D.; Muench, V.; Stynes, J., “Mechanisms of component-
                   oriented software development“, Internet Research: Electronic
                   Networking Applications and Policy, Volume 9 No.1, 1999

[HPWSMF]           Web Services Management Framework

[Huhns 02]         Huhns, Michael N., “Agents as web services”,
                   IEEE Internet Computing¸ Issue July / August, 2002

[Interoute]        “What is a Virtual Private Network?”, Interoute Ltd.

[ITIL]             Office of Government Commerce – IT Infrastructure Library (ITIL)

[ITSMF]            IT Service Management Forum
                   “Introduction to IT Service Management”,

[Jaruzelski 00]    Jaruzelski, B.; Ribeiro F.; Lake, R.; “ASP 101: Understanding the
                   Application Service Provider Model”, Booz-Allen & Hamilton, 2000

[Jerskey 02]       Jerskey, Pamela, “Top Security Risks in the Future”, Audit News Article,
                   Boston College, 2002

[Jini]             Jini Network Technology

[Kariyawasam 01] Kariyawasam, Rohan, “Telecoms Regulation: Peering and Transit Over
                 TCP/IP Networks”,
                 Computer Law & Security Report, Vol. 17 No. 1, 2001

[Kashyap 01]       Kashyap, Rajiv, “The Effects of Service Guarantees on External and
                   Internal Markets”,
                   Academy of Marketing Science Review, Vol. 2001 No. 10, 2001

[Kenyon 03]        Kenyon, C.; Cheliotis, G., “Grid Resource Commercialization: Economic
                   Engineering and Delivery Scenarios“, in Grid Resource Management:
                   State of the Art and Research Issues, Kluwer, 2003

[Klein 83]         Klein, L., “The Economics of Supply and Demand”,
                   Blackwell Publishers, 1983

[Kreger 03]        Kreger, Heather, “Fulfilling the Web Services Promise”,
                   Communications of the ACM, Vol. 46 No. 6, 2003

[Kubiatowicz 03]   Kubiatowicz, John, “Extracting Guarantees from Chaos”,
                   Communications of the ACM, Vol. 46 No. 2, 2003

[Lande 00]         Lande, L., “Software for rent: region is ripe for a new kind of online
                   market”, Asian Wall Street Journal Weekly, Vol. 22 No. 18, 2000

[Lange 99]         Lange, D.B.; Oshima, M., “Seven Good Reasons for Mobile Agents”,
                   Communications of the ACM, Vol. 42 No. 3, 1999

[Lau 03]           Lau, H.C.W.; Wong, C.W.Y.; Ngai, E.W.T.; Hui, I.K., “Quality
                   management framework for a virtual enterprise network: a multi-agent
                   approach”, Managing Service Quality, Vol. 13 No. 4, 2003

[Lee 98]           Lee, Ho Geun, “Do Electronic Marketplaces Lower the Price of Goods?”,
                   Communications of the ACM, Vol. 41 No. 1, 1998

[Lee 03]           Lee, J.-N.; Huynh, M.Q.; Kwok, R.C.-W.; Pi, S.-M.,”IT Outsourcing
                   Evolution – Past, Present, and Future”,
                   Communications of the ACM, Vol. 46 No.5, 2003

[Legion]           Legion – Worldwide Virtual Computer

[MAC]              Technology – Glossary, MAC Telecom S.A.

[MacKie-Mason 97] MacKie-Mason, J.; Murphy L.; and Murphy J., “Responsive Pricing in
                  the Internet”, in Internet Economics, ed. McKnight and Bailey,
                  MIT Press, 1997

[Macskasi 02]      Macskasi, Zsolt, “Customer Lock-In with Long-Term Contracts”,
                   Northwestern University, 2002

[McKnight 00]      McKnight, L.W.; Boroumand J., “Pricing Internet services: after flat
                   rate” Telecommunications Policy, No. 24, 2000

[McPhillips 99]   McPhillips, Elizabeth “The Structure and Trends of the ISP Market”,
                  University of Strathclyde, Glasgow and Hewlett-Packard Laboratories,
                  Bristol, 1999

[Meng]            Meng, J.; Helal, S.; Su, S., “An Ad-Hoc Workflow System Architecture
                  Based on Mobile Agents and Rule-Based Processing”, University of

[Mennie 00]       Mennie, David William, “An Architecture to Support Dynamic
                  Composition of Service Components and its Applicability to Internet
                  Security”, Master Thesis, Carleton University, Ottawa, Canada, 2000

[Mertens 98]      Mertens, P; Griese, J.; Ehrenberg, D., “Virtuelle Unternehmen und
                  Informationsverarbeitung“, Springer Verlag, 1998

[Metz 03]         Metz, Cade, “Grid Computing: Case Study – Entelos”, PC Magazine,
                  January 1, 2003

[Milojicic 02]    Milojicic, D.S.; Kalogeraki, V.; Lukose, R.; Nagaraja1, K.; Pruyne, J.;
                  Richard, B.; Rollins, S.; Xu, Z., “Peer-to-Peer Computing“,
                  Hewlett- Packard Company, 2002

[MSTechnet]       Microsoft Technet – Services Overview

[Mundt 02]        Mundt, Thomas, “Spontane Verknüpfung von Diensten –
                  Wiederverwendung und Kostenersparnis“,
                  Dissertation, University of Rostock, 2002

[MWOL]            Merriam-Webster Online Dictionary

[NET]             Microsoft .NET

[NIST 01]         “Risk Management Guide for Information Technology Systems”,
                  Recommendations of the National Institute of Standards and Technology,
                  Special Publication 800-30, U.S. Department of Commerce, 2001

[NIST 03]         “Guide to Information Technology Security Services”,
                  Recommendations of the National Institute of Standards and Technology,
                  Special Publication 800-35, U.S. Department of Commerce, 2003

[NorduGrid]       The NorduGrid Project

[Normann87]       Normann, R., “Dienstleistungsunternehmen”, Hamburg 1987

[Odlyzko 98]      Odlyzko, Andrew, ”Paris Metro Pricing for the Internet “, ACM
                  Conference on Electronic Commerce, 1998

[Odlyzko 01]      Odlyzko, Andrew, ”Internet pricing and the history of communications”,
                  AT&T Labs Research, 2001

[OECD]            Organisation for Economic Co-operation and Development

[OMG]             Object Management Group

[OGSA]            Global Grid Forum, “Open Grid Services Architecture”

[Oleson 98]       Oleson, Thomas D., “Price of Precision”,
                  CIO Magazine, February 15 Issue, 1998

[Parameswaran 01] Parameswaran, M.; Susarla, A.; Whinston, A.B., “P2P Networking: An
                  Information-Sharing Alternative”, IEEE Computer, July Issue, 2001

[Parish 97]       Parish, Robert J., “Service level agreements as a contributor to TQM
                  goals”, Logistics Information Management, Vol. 10 No. 6, 1997

[Paul 00]         Paul, Lauren, “The Need for Speed”, Network World, February 7, 2000

[Pender 00]       Pender, Lee, “How to Communicate Value”,
                  CIO Magazine, December 15 Issue, 2000

[Peppard 03]      Peppard, Joe, “Managing IT as a Portfolio of Services”,
                  European Management Journal, Vol.21 No. 4, 2003

[PKI]             The PKI Page

[Porter 80]       Michael E Porter, “Competitive Strategy”, Free Press, 1980

[PUMA 98]         P ublic Management Service, “Best Practice Guidelines for User
                  Charging for Government Services”, OECD, 1998

[Quan 02]         Quan, Daniel C., “The Price of a Reservation”, Cornell Hotel and
                  Restaurant Administration Quarterly, June Issue, 2002

[Rao 00]           Rao, Siriginidi Subba, “Internet Service Providers in India”, Work Study,
                   Vol. 49 No. 5, MCB University Press, 2000

[Ratza 93]         Ratza, C.L., “A client driven model for service pricing”, Journal of
                   Professional Services Marketing, Vol. 8 No. 2, 1993

[Rowley 97]        Rowley, Jennifer, “Principles of price and pricing policy for the
                   information marketplace”, Library Review, Vol. 46 No. 3, 1997

[Rust 03]          Rust, Roland T., “E-Service: A new paradigm for the Business in the
                   electronic environment”,
                   Communications of the ACM, Vol. 46 No. 6, 2003

[Sairamesh 98]     Sairamesh, J.; Codella, C., “NetBazaar: Networked Electronic Markets
                   for Trading Computation and Information Services”, Proceedings of the
                   7th Delos Workshop on Electronic Commerce, Springer Verlag, 1998

[Schmalensee 84]   Schmalensee, R., “Gaussian demand and commodity bundling”,
                   Journal of Business,Vol. 57, 1984

[Seth 03]          Seth, Manoj, “Introduction to Web Services Management”,
         , February 12, 2003

[Shapiro 98]       Shapiro, C.; Varian, H.R., “Information Rules”,
                   McGraw-Hill Publ., 1998

[Sheizaf]          Sheizaf, Rafaeli, “Different Auction Models”, web page

[Sharma 02]        Sharma, S.K.; Gupta, J. N.D., “Application Service Providers: Issues and
                   Challenges”, Logistics Information Management, Vol. 15 No. 3, 2002

[Sieber 99]        Sieber, P.; Griese, J., “Organizational Virtualness and Electronic
                   Commerce”, Proceedings of the 2nd International VoNet – Workshop,
                   Simowa Verlag, 1999

[Spahni 02]        Spahni, D.; Meier, J., “Web Services im eGovernment: Vision und
                   Konzept veränderter Wertschöpfungsketten der staatlichen
                   Leistungserbringung“, Berner Fachhochschule, Switzerland, 2002

[Stevens 02]       Stevens, Michael, “Service-Oriented Architecture Introduction, Part 1”,
         , 2002

[Stafford 03]      Stafford, Thomas M., “E-Services”,
                   Communications of the ACM, Vol. 46 No. 6, 2003

[Stiller 98]       Stiller, B.; Fankhauser G.; Plattner B.; Weiler N., “Charging and
                   Accounting for Integrated Internet Services – State of the Art, Problems,
                   and Trends”, Proceedings of INET ’98: The Internet Summit, 1998

[Stiller 01]       Stiller, B.; Reichl, P.; Leinen, S., “Pricing and Cost Recovery for Internet
                   Services: Practical Review, Classification, and Application of Relevant
                   Models“, Netnomics: Economic Research and Electronic Networking,
                   Vol. 3 No. 2, 2001

[Sun 02]           Sun Microsystems Inc., “Implementing Effective Chargeback Policies”,
                   White Paper, 2002

[Sundarajan 03]    Sundarajan, Arun, “Nonlinear pricing of information goods”, Leonard N.
                   Stern School of Business, New York University, 2003

[Tanner 03]        Tanner, John, “Grid computing set for big growth”, America’s Network,
                   Vol. 107 No. 8, 2003

[TeraGrid]         The TeraGrid Project

[Tsalgatidou 02]   Tsalgatidou, A.; Pilioura, T., “An overview of standards and related
                   technology in web services”,
                   Distributed and Parallel Databases, No. 12, 2002

[Tung 97]          Tung, W.; Capella, L.M.; Tat, P.K., “Service pricing: a multi-step
                   synthetic approach”,
                   The Journal of Services Marketing, Vol. 11 No. 1, 1997

[Turisco 00]       Turisco, Fran, “Evaluating ASP Pricing Models and Contract Terms”,
                   Healthcare Financial Management¸ August Issue, 2000

[Tutor2u]          “Pricing – Influences on Pricing Policy”,

[W3CWS]            W3C Web Services Architecture Working Group, Web Services
                   Architecture Requirements, W3C Working Draft (Version: August 8,

[Webopedia]        Webopedia – Online Dictionary for Computer and Internet Terms

[Wehrli 01]        Wehrli, Hans-Peter, “Marketingpraxis“ CD-Rom Student Edition, Haufe
                   und Kisling Verlag AG, Universität Zürich, 2001

[Wittmann 00]      Wittmann, A., “Service providers and outsourcing”,
                   Network Computing, Vol. 11 No. 25, 2000

[Wladawsky 02]      Wladawsky-Berger, Irving, “The Future Utility of IT”,
                    Optimize, November Issue, 2002

[Wolski 03]         Wolski, R.; Brevik, J.; Plank, J.S.; Bryan, T., ”Grid resource allocation
                    and control using computational economies”, Grid Computing – Making
                    the Global Infrastructure a Reality, John Wiley & Sons Ltd, London,

[Zapthink 03]       ZapThink, LLC, “Solving the IT impasse with service orientation”,
                    Whitepaper, 2003

[Zigmont 99]        Zigmont, Jason, “Pricing your Internet Services”,, 1999

All Internet links last visited on December 29, 2003.