A Model for Web Services Discovery with QoS

Document Sample
A Model for Web Services Discovery with QoS Powered By Docstoc
					                      A Model for Web Services Discovery with QoS
                                         Paper by Shuping Ran
               Critique Report by Eyhab Al-Masri (Discussant) and Razieh Niazi (Presenter)

Brief summary
         In the paper titled “A Model for Web Services Discovery with QoS” written by Shuping Ran, the
author proposes a solution to enhance the discovery of Web services using Quality of Service (QoS)
attributes. In recent years, QoS support for Web services has become a widely researched area and has
shown to be an effective mechanism in Web services’ discovery particularly in differentiating between
services that share similar functionalities. In the proposed solution, Ran recommends a new discovery
model for Web services with the addition of QoS Certifier which is responsible for verifying QoS claims
supplied by service providers and issuing Certification IDs for services whose QoS claims can be verified.
         In order to incorporate QoS Certifier into the existing Web service discovery model, Ran
introduces a slight modification to the current Web services’ architecture such that registration is performed
at the UDDI only after a service provider verifies QoS claims with the Certifier. Under this new model, a
service provider prepares a set of QoS measurements and communicates this information to the Certifier
which then begins checking QoS claims and sends back a notification to the service provider. This
notification is either an acknowledgement that the QoS claims have failed, or a Certification ID that the
QoS claims have been successfully verified. Once the service provider is issued a Certification ID, it begins
the registration process through the UDDI registry, providing both functional and non-functional
information (QoS claims), along with the Certification ID. The UDDI performs a verification process for
the Certification ID and communicates with QoS Certifier to check whether it is valid or invalid. Once
UDDI verifies this information, it registers the service provider’s information.

Points in favor
         The framework proposed by Ran offers an enhanced Web services’ discovery model in the sense
that it provides a mechanism for service providers to associate and include QoS metrics within the UDDI
registry. In addition to that, verifying QoS information is applied by QoS Certifier which gives assurances
to clients that supplied QoS are valid and can be trusted. Furthermore, the solution allows clients to
articulate their search queries based on QoS attributes. The paper also suggests at least twenty QoS
properties that can be measured and associated with Web services.

Points Against
         Although Ran’s solution has the potential of providing QoS support for Web services’ discovery,
there are technical challenges and limitations associated with the proposed system. One of these limitations
is the redundancy of performing QoS measurements. QoS measurements have to be supplied by the service
provider at the time of registration while QoS Certifier will eventually perform another round of QoS
measurements to verify the service provider’s claims, and therefore, QoS measurements are conducted at
least twice. In addition to that, Ran’s solution proposes a major change to the existing UDDI specification
(Version 3), which may be very problematical at this stage due to the fact that many software vendors have
already endorsed the UDDI standard with its latest specification and have incorporated UDDI support
within their software products such as Enterprise Windows Server 2003, WebSphere, and among others.
         Ran’s paper did not provide a clear explanation on key issues such as the storage of QoS
information within qualityInformation tModel, how data types will be managed or declared, and the
possibility of linking QoS information to bindingTemplates. The ability to incorporate or relating QoS to
bindingTemplates is very useful in the sense that they can be used for multiple offerings of the same Web
service. For instance, a service provider that has variations of the same Web service (i.e. Enterprise,
Professional, and Basic) will eventually have different QoS measurements for each of these offerings, and
therefore, applying the same QoS information to all bindingTemplates may not be accurate or helpful.
         In addition to that, Ran’s solution lacks implementation details or experimental results. The system
proposed only provides an idea of what can be done, however, no design, implementation, or performance
testing could be found anywhere in the paper. The author only provides one example of a SOAP
request/response for illustration purposes, and does not provide any details on how the system can be
implemented or designed. Potentially, the author could have implemented a working model for the twenty
Web service attributes and how they can be used within the UDDI through existing tModels structure.
However, the lack of implementation details and experimental results makes the task of verifying the
feasibility of this solution very difficult.
         Furthermore, having QoS information at the time of registration will impose a restriction on service
providers in the sense that they have to continually provide QoS claims with every Web service that needs
to be published as the author suggests, “In the proposed model, a Web service provider needs to supply
information about the company …. as well as to supply quality of service information…” (Section 2.2).
The fact that every time a service provider has to provide QoS claims at the time of registration may not be
potentially suitable due to the fact that service providers have to devote more time, money, and extra effort
to perform QoS measurements on their Web services (i.e. use built-in or third party QoS measurement
tools). In addition, requiring service providers to supply QoS information at the time of registration may
allow them to think of ways on how to improve QoS measurements.
         Another major limitation found in Ran’s paper is periodic updates of QoS information. Due to the
nature of how Web services operate, and how they strictly depend on network traffic, it becomes critical to
perform periodic updates to QoS measurements to ensure that the claims are constantly valid.
         Many of the QoS metrics are performed over a period of time (i.e. days, weeks, months) which
makes the task of conducting QoS measurements very time consuming. However, from Ran’s paper, it
conveys that QoS measurements are semi-instant or simple to generate which is not entirely true. In
addition to that, in Ran’s solution the UDDI is required to perform additional tasks such as verifying
Certification IDs at the time of registration as the author suggests, “UDDI communicates with the certifier
to check the existence of the certification…” (Section 2.2). Performing this task will likely increase the
overhead and work executed by the UDDI which may not be practical since UDDI is intended to serve as a
service directory and does not have the capability to serve as a communication link. Furthermore, the
UDDI does not have the built-in functionality to communicate with third parties, and the paper does not
clearly define how this task can be performed or developed.

Comments for Improvement & Conclusion
         Although Ran’s solution provides a way to include QoS information at the UDDI registry, there is
room for improvement. The supply of QoS information should be made optional and service providers
should not be enforced to provide them at anytime. In fact, supplying this information by service providers
will necessitate the verification of their QoS claims (i.e. Certifier); however, it would be desirable if there is
a service broker that can collect QoS information about Web services in a transparent manner based on
entries within the UDDI registries. This will not force service providers to devote efforts on providing QoS
measurements at the time of registration and prevents any room for manipulation. In addition, having a
service broker for measuring QoS information for Web services does not require any modifications to be
made to existing standards (i.e. UDDI) and enables the monitoring of Web services across stages (i.e.
traffic analysis, reputation, and among many others).
         Ran’s solution provides a mechanism for improving the discovery of Web services based on QoS
attributes, however, it lacks implementation results, and requires major changes to the existing UDDI
standard. In order to improve this system, it is important to consider the fact that QoS measurements should
be conducted in a fair and transparent manner, clients should be able to customize the discovery process
using a combination of QoS parameters as constraints, and support of periodic QoS updates is unavoidable.

Ran, S., “A Model for Web Services Discovery with QoS”, ACM SIGecom Exchanges 4(1): pp. 1-10, 2003