Community Computing Framework for Enhancing Internet Services by djy18697


									EurAsia-ICT 2002, Shiraz-Iran, 29-31 Oct.

                  Community Computing Framework for Enhancing Internet Services

                                         Byung Y. Sung, Behrooz Shirazi, and Mohan Kumar
                                         Department of Computer Science and Engineering,
                                                  University of Texas at Arlington,
                                          P.O. Box 19015, Arlington, TX 76019-0015, USA
                                                {sung, shirazi, kumar}

                                  Abstract                                 Next Generation Internet initiatives [10].
                                                                               PICO deals with the creation of mission-oriented dy-
            With an objective to enhance the existing Internet             namic computing communities that perform tasks on be-
         computing model, we are developing a general-purpose              half of users and devices autonomously. PICO consists of
         framework called Pervasive Information Community Or-              software entities called delegents (intelligent delegates) and
         ganization (PICO). PICO deals with the creation of                hardware devices, called camileuns (connected, adaptive,
         mission-oriented dynamic computing communities that per-          mobile, intelligent, learned, efficient, ubiquitous nodes).
         form tasks on behalf of users and devices autonomously.           PICO encompasses computing, communication, and dy-
         PICO‘s communities are made up of software agents                 namic context-aware information processing [7, 8, 9].
         called delegents(intelligent delegates). Community com-               For example, in a PICO-enabled office management en-
         puting will enable provisioning of context-aware, location-       vironment a CEO can set up a meeting by simply specifying
         aware, just-in-time services all the time everywhere. PICO-       the meeting time. The CEO’s delegents will automatically
         enabled pervasive environments are expected to revolution-        form a community with the delegents of all involved par-
         ize the way Internet is employed, enhance quality of life and     ties to schedule the meeting. The delegents in the meeting
         promote user-centric applications. In this paper we discuss       community use the participants’ calendar information to re-
         formal models for delegents, delegent communities and the         solve conflicts and determine the mode of participation by
         methodology for creation and destruction of communities.          different employees. Some of the employees will attend the
                                                                           meeting in person, some will participate via video confer-
                                                                           encing, yet some will use teleconferencing to participate.
         1. Introduction                                                       The concept of PICO extends the current notion of perva-
                                                                           sive computing; namely, that computers are everywhere [3].
             In the very near future, the Internet is expected to be em-   Pervasive computing is already in use partially: in pocket
         ployed to facilitate such applications as telemedicine, edu-      PCs, mobile phones, laptops, desktops, watches, and other
         cation, transportation, and military all the time and every-      devices. With pervasive computing alone, we are also often
         where. Advances in computing and Internet technologies            inundated with information - out of context, out of date, or
         have significantly influenced industry, education, business,        unwanted. The objective of PICO is to provide automated
         and commerce. Additionally, the way in which informa-             continual services to improve quality of life. The novelty of
         tion is gathered, processed, and disseminated will play a vi-     the PICO initiative lies in creating communities of delegents
         tal role in health, education, transportation, military, polic-   that collaborate proactively to handle dynamic information,
         ing, entertainment, and lifestyle in general. These appli-        provide selective content delivery, and network availabil-
         cations demand automated, continual unobtrusive services          ity. More precisely, PICO provides what we want, when
         and proactive real-time collaborations among devices and          we want, where we want and how we want type of services
         software agents in dynamic heterogeneous environments.            autonomously, unobtrusively, and continually.
         The existing push-/pull-based Internet and network model              The next section presents an overview of on going re-
         which is intended essentially for static information services     search initiatives on pervasive computing as the background
         is inadequate to meet these requirements. To provide the in-      work of the PICO framework. The Section 3 describes the
         frastructure needed by such applications, we propose Perva-       components of the PICO framework and discuss how they
         sive Information Community Organization (PICO) - a mid-           are put together to support the PICO community comput-
         dleware framework to enhance the existing Internet model.         ing paradigm. The section 4 describe how a community is
         The PICO framework will complement Internet2 [5] and              formed and managed. It also describes the effectiveness of

Page 177
      the community computing framework. Finally, the Section          and automatically to suit any application and/or user re-
      5 concludes the paper.                                           quirements. PICO differs from all the existing projects we
                                                                       are aware of. In PICO, delegents constantly learn about a
      2. Pervasive Computing Research Initiatives                      person (device or the environment), work diligently to as-
                                                                       sist the person in every task in an unobtrusive way. Com-
                                                                       munities are dynamically formed around a user to provide
          The pervasive computing idea was initially put forward       needed services transparently.
      by Mark Weiser [16]. Recently, pervasive computing has
      been an active area of research attracting substantial atten-
      tion by several research groups. In situated computing [15],     3. The PICO Framework
      situations of use and context play a central role in the use
      of computers. Interfacing mobile devices and environment-            As we mentioned earlier, PICO deals with the creation
      based applications are the main themes in situated comput-       of mission-oriented dynamic computing communities that
      ing and sentient computing [2] paradigms which are both          perform tasks on behalf of users and devices autonomously.
      dependent on sensory inputs from the environment. The            Camileuns and delegents are the basic building blocks of
      concept of sentient computing is to create an interface that     PICO. We argue that the novelty of the PICO initiative
      extends throughout the environment. The contextual com-          lies in creating communities of delegents that collaborate
      puting paradigm [11] at Georgia Tech employs video inputs        proactively to handle dynamic information, provide selec-
      for location tracking and management. The Oxygen project         tive content delivery, and network availability. In this sec-
      [3] at MIT focuses on eight environment-enabled technolo-        tion, we first describe camileuns and delegents in detail, and
      gies: H21, E21, N21 and five others aimed at improving the        then show how a set of camileuns and delegents are mapped
      user experiences such as speech, intelligent knowledge ac-       to a mission-oriented community. The details of PICO can
      cess, collaboration, automation of everyday tasks and adap-      be found in [8, 9].
      tation of machines to the user needs [3]. The Oxygen
      project is a large all encompassing research initiative whose    3.1. Camileuns
      current emphasis is on improving human-computer inter-
      face, PICO’s objective is to enhance the existing, as well          In the foreseeable future, the environment will be replete
      as future, Internet services to improve human quality of life    with adaptable smart devices, called camileuns. Typically, a
      through pervasive computing.                                     camileun possesses one or more of the functionalities such
          The Portalano project [4] at the University of Wash-         as: see, hear, adapt, compute, communicate, learn, and pro-
      ington seeks to create a test-bed for investigation into the     cess information. Camileuns can be of different types and
      emerging field of Invisible Computing. Portalano’s empha-         complexities—a temperature sensing device, an active net-
      sis is on data-centric routing to facilitate automatic ’smart’   work node, and a state-of-the-art workstation. They can
      data migration among applications rather than Internet ser-      communicate in broadband, wired/wireless, or optical net-
      vice enhancement. The Aura project at CMU is another             works. Camileuns are self-adaptive in the sense they can
      exciting pervasive computing project which encompasses           change their behavior and functionalities either to suit the
      techniques and concepts such as task-driven computing, re-       environment they are in, or the application tasks they are
      source exploitation, the use of agents [14], and ubiquitous      required to perform.
      service provisioning. Aura is based on well-proven legacy           For modelling purpose, a camileun can be described
      systems such as Odessey [13] and Coda [12]. Researchers          by C =< Cid , S, F >, where Cid is the identifier
      at IBM are developing a model for writing application front-     of the camileun, S is the system characteristics of the
      ends for their Platform-Independent Model for Applications       camileun, and F is the set of functions it is expected to per-
      (PIMA) project [1]. The PIMA environment allows a devel-         form. For example, a camileun mounted on a street lamp
      oper to create platform-independent front-ends to multiple       shown in the Figure 2 can be described by Clamp =<
      heterogeneous device platforms. In the Endeavour project         Clamp , Slamp , Flamp > with the system characteristics,
      at the University of California Berkeley, scalable adaptable     Slamp =< operating system; processor type; memory; I/O
      information utility is the key issue [6]. Nomadic data access    type; camera; router; wireless transceiver; . . . > and the
      is facilitated by pervasive storage of information in the net    functionality set Flamp =< surveillance; traffic monitor-
      to enable users to acquire information anywhere anytime.         ing; processing; communicating; . . . >.
      Unlike proactive delegent communities in PICO, the com-
      ponents of Endeavor react to changes in the environment.         3.2. Delegents
          PICO’s uniqueness, simplicity and versatility are derived
      from proactive community computing and modular hard-                A delegent is an intelligent delegate that works diligently
      ware and software tools that can be assembled dynamically        on behalf of a camileun or a user. For example, a dele-

Page 178
      gent can gather information locally or remotely, collaborate
      with other delegents to form a computing community. A
      delegent works in a community environment where it inter-
      acts with other delegents and environments. The delegent
      responses to sensory inputs, events in the community and
      events within itself, and takes appropriate actions based on
      a set of rules.
          A delegent is represented by a two tuple, Delegent =<
      Did , Fd > where Did is the delegent’s identity and Fd is its
      functional description. The identity of a delegent is given by
      the 3-tuple: Did =< id, Cid , P >, where id is the delegent
      identifier, Ci d is the camileun where it was created, and P         Figure 1. An example of camileun-delegent as-
      is the community it belongs to. Functionally, a delegent is         sociations.
      described by a three tuple: Fd =< M, R, L >, where M is
      the set of program modules, R is the set of rules for building
      a delegent, and L is the set of low level functions. For ex-
      ample, a delegent associated with the street lamp camileun
      shown in the Figure 2, D(streetlamp, camera) can be de-
      fined as, D1 =< M1 , R1 , L1 >, where M1 =< image cap-
      ture mechanism; event detector; timer; . . . >, R1 =< state
      transition; migration; new connection; . . . >, and L1 =<
      pattern recognition; classification . . . >.

      3.3. Communities

          A community of PICO consists of one or more dele-
      gents working towards achieving a common goal. Com-
      munity formation and dismantling is a rule-based process.              Figure 2. PICO as an middleware services
      The rules for the formation and dismantling are defined in
          A community is defined by P =< D, φ >, where D                ployment of delegents. Mobile delegents interact with a
      is the set of delegents, and φ is the community’s mission        delegent manager to provide community-based services. A
      or goal. As an example, community for the PICO-enabled           community is a platform for a set of delegents to collabo-
      office environment described in the section 1, can be de-         rate seamlessly and efficiently to allow the provisioning of
      fined as P1 =< D1 , φ1 >, where D1 =< D(pocketP C);               combined services in seemingly a single point of contact.
      D(car); D(desktop); D(cellphone). . . >, D(x), is a                 Most existing software agent systems assume that the
      delegent representing camileun x in this context and φ1 =<       agent execution environment is homogeneous. The assump-
      schedule the meeting >. Further detailed discussions on          tion on homogeneous agent execution environment is no
      community related issues in PICO framework, refer to the         longer valid in pervasive computing where thousands of dif-
      following section 4.                                             ferent types of devices coexist. In the PICO framework,
                                                                       camileuns are abstracted representations of computing de-
         The PICO middleware services shown in the Figure 2 il-        vices, and delegents are mobile software agents. In the
      lustrates the relationship among camileun, middleware ser-       PICO concept, we are able to formulate a way of supporting
      vices provided by delegents and communities, and appli-          software agents in heterogenous execution environment.
      cations. The PICO middleware services are distributed               Further, the existing software agent systems do not dif-
      over camileuns that possess function capability to provide       ferentiate device functionalities since they assume that all
      required services. The PICO framework provides seam-             devices provide the same set of services. In the PICO frame-
      less interaction between applications and existing comput-       work, camileun-delegent associations enable camileuns to
      ing/communication infrastructures include the internet.          behave like the real chameleons. A camileun can be repre-
         In the PICO design, the PICO middleware services are          sented by one or more delegents, each of which provides a
      realized by a special delegent called delegent manager. The      different service. For example, as shown in the Fig. 1, the
      delegent manager performs management tasks including             street lamp camileun can serve as a network router, a video
      process, memory, mobility management, creation and de-           surveillance camera, or an information kiosk. The delegent,

Page 179
      in the middle of the Figure 1 is associated with the street       community. The tasks carried out by the delegents are coor-
      lamp to provide wireless network services to those who are        dinated to make sure that every task is carried out in order,
      within the transmission range. Another delegent can be as-        since some tasks are dependent on others.
      sociated with the street lamp to monitor vehicular traffic.
                                                                        4.2. The PICO community operations
      4. Community Computing in PICO
                                                                            As we discussed earlier, the major architectural issues
                                                                        for realization of the PICO community concept are 1) when
         Communities of PICO consist of one or more delegents
                                                                        and how to form communities, and 2) how to manage the
      working to achieve a common goal. The major architec-
                                                                        communities. The community management issue of the
      tural issues for realizing the PICO community concept are
                                                                        PICO framework will be discussed in this section.
      1) when and how to form communities, and 2) how to man-
      age the communities. Communities are the highest level of             The formation and dismantling of a community is a rule
      interactions that require delegents to interact each other fre-   based operation. The community formation rule is defined
      quently, continuously, and actively. In the next section, we      in delegents based on the characteristics and the level of
      will discuss the characteristics of four levels of interactions   interaction associated with the events. An event is sensed
      in the PICO framework. The community management is-               or generated by a delegent. Like most software agents, the
      sue of the PICO framework is addressed and discussed in           delegent receiving an event reacts to it based on the prede-
      the section 4.2.                                                  fined rules. The predefined rules may require the delegent
                                                                        to create a community.
                                                                            The functionalities of a delegent are defined by D =<
      4.1. Levels of interactions in the PICO framework                 M, R, L >. The tuple R represents a set of rules for the
                                                                        delegent. The rules can be either statically defined by the
          The PICO framework defines four levels of interactions         designer or dynamically formulated by using agent learning
      between delegents: communication, cooperation, collabo-           techniques. Currently, we assume that the set of rules for a
      ration, and community. The characteristics of these interac-      delegent is defined a priori.
      tion levels are summarized in the Table 1.                            The rules for forming a community are largely deter-
          Communication level interaction is a basic form of in-        mined by the level of interaction among delegents work-
      teraction in that delegent are loosely coupled and only ex-       ing together to carry out tasks associated with the event.
      change data. The duration of communication level inter-           Generally, an event required to form a community requires
      action is relatively short. An example of this level of in-       continuous interactions among participating delegents. The
      teraction is that one delegent passes a short notice to an-       delegents in a community strongly depend on each other
      other delegent which doesn’t entail further exchange of           and heavily participate to handle the events as described in
      messages. Cooperation level interaction is almost identi-         the section 4.1.
      cal to the communication level except that the duration of            The other factor to determine community formation is
      interaction is long and continuous. The cooperation type of       the characteristics of tasks needed to process the event. The
      interaction occurs in sensor networks where a sensor con-         transactional property of an event triggers formation of a
      tinuously transmits its data to the recipients.                   community, and it requires delegents to finish their assigned
          Collaboration and community level interactions are more       tasks completely and successfully. In the example of the
      stronger forms of interactions than those of communica-           PICO-enabled office management environment, described
      tion and cooperation level. The collaboration and commu-          in the section 1, the CEO’s delegent has to ensure that meet-
      nity level interactions not only exchange data but also allow     ing arrangement is completed without any ambiguity on
      member delegents to share control privilege within the com-       time, place, and mode of the meeting.
      munity.                                                               The delegent that receives the event initiates the forma-
          The purpose of defining the four levels of interactions        tion of a community with the help of PICO middleware
      is to identify the characteristics of the PICO’s community        services. Once a community is formed, it will be main-
      computing. It also help us to build rules for community           tained by the PICO middleware services and the initiating
      formation which can be defined in delegents.                       delegent. The PICO middleware services perform manage-
          Unlike software agents in homogeneous execution en-           ment tasks including mobility management and application
      vironment, delegents may execute on might be running on           level message routing for delegents. Currently, the commu-
      heterogeneous devices, the camileuns. The camileun can be         nity membership information is maintained by the initiating
      heterogeneous in term of functionality and computing capa-        delegent but we are investigating efficient ways of distribut-
      bility. The delegents are also required to carry out a set of     ing the membership information to reduce the workload of
      tasks based on their functionality to achieve the goal of the     the initiating delegent and provide fault-tolerance.

Page 180
                      Table 1. Summary of characteristics of four interaction levels in the PICO framework.
                                              Communication Cooperation Collaboration Community
                      Mode of Interaction            Data                Data           Data,Control       Data,Control
                      Duration of Interaction        Short               Long           Short              Long
                      Frequency of Interaction       Occasional          Medium         Medium             High
                      Degree of Dependency           N/A                 N/A            Weak               Strong
                      Degree of Participation        N/A                 N/A            Medium             High
                      Number of Shared Tasks         N/A                 N/A            Single             Multiple

      5. Conclusion                                                        [6] R. H. Katz.         The endeavour expedition:          Chart-
                                                                               ing the fluid information utility, web/internet site:
          PICO’s novelty lies in the use of community computing            [7] M. Kumar and S. K. Das. Flexible integrated cache for effi-
      concept and the association of camileuns and delegents. In               cient information access in mobile computing environments.
      this paper, we discussed formal models for delegent, dele-               In NSF Workshop on an Infrastructure for Mobile and Wire-
      gent communities and the methodology for creation and de-                less Systems, Scottsdale, AZ, 2001.
      struction of communities. We also argued how the PICO                [8] M. Kumar, S. K. Das, B. Shirazi, D. Levine, J. Marquis, and
      community computing differs from the the existing soft-                  L. Welch. Pervasive information community organization
      ware agent systems. Unlike most existing researches on                   (pico) of camileuns and delegents for future global network-
      software agent systems, the PICO framework is focused on                 ing. In LSN Workshop, Vienna, USA, 2001.
                                                                           [9] M. Kumar, S. K. Das, B. Shirazi, B. Y. Sung, and D. Levine.
      solving issues and challenges in pervasive computing in-
                                                                               Pervasive information community organization (pico): A
      cluding software adaptability, efficient agent coordination               framework for enhancing internet services. Technical report,
      and collaboration, agent mobility management, application                2001.
      message routing for agent. The PICO framework provides              [10] NGI.             Next    generation     internet   initiative,
      simple, yet effective abstraction of various device which can  
      be used to overcome software adaptability. The PICO com-            [11] W. Rungsarityotin and T. Starner. Finding location using
      munity computing provides more elegant way of collabo-                   omni directional video on a wearable computing platform.
      ration and coordination to pervasive computing. Several                  In IEEE International Symposium on Wearable Computing
      other issues and challenges of the PICO frame such as self-              (ISWC2000), Atlanta GA, 2000.
                                                                          [12] M. Satyanarayanan. Coda: A highly available file system for
      organization of communities and fault-tolerance will be the
                                                                               a distributed workstation environment. In the Second IEEE
      future works.                                                            Workshop on Workstation Operating System, Pacific Grove,
                                                                               CA, 1989.
                                                                          [13] M. Satyanarayanan. Mobile information access. IEEE Per-
                                                                               sonal Communications, 3(1), 1996.
                                                                          [14] M. Satyanarayanan. Pervasive computing: Vision and chal-
       [1] G. Banavar, J. Beck, E. Gluzberg, J. Munson, J. Sussman,            lenges. IEEE Personal Communications, 8(4):10–17, 2001.
           and D. Zukowski. Challenges: An application model for per-     [15] A. Schmidt, W. Van de Velde, and G. Kortuem. Situated
           vasive computing. In 6th Annual International Conference            interaction in ubiquitous computing. In Workshop on ”Sit-
           on Mobile Computing and Networking (MOBICOM 2000),                  uated Interaction in Ubiquitous Computing” at CHI 2000,
           pages 266–274, Boston MA, USA, 2000.                                2000.
       [2] R. Curwen, A. Hopper, P. Steggles, and A. Ward. Sentient       [16] M. Weiser. The computer for the 21st century. Scientific
           computing. (video), AT&T Laboratories Cambridge, 1999.              American, 265(3):94–104, 1991.
       [3] M. L. Dertouzos. The future of computing. Scientific Amer-
           ican, 281(2):52–55, 1999.
       [4] M. Esler, J. Hightower, T. Anderson, and G. Borriello. Next
           century challenges: data-centric networking for invisible
           computing. the portolano project at the university of wash-
           ington. In Fifth Annual ACM/IEEE International Confer-
           ence on Mobile Computing and Networking. ACM, pages
           256–62, Seattle WA, USA, 1999.
       [5] Internet2.           Internet2:       Internet2    website,

Page 181

To top