Dottorato di Ricerca in Ingegneria dell Informazione e sua applicazione nell Industria e nei Servizi Integration and Coordination in in both Mediator-Based and Peer-to-Peer Systems presenter: (pense@inform.unian.it) (http://www.inform.unian.it/~pense/pense.htm) Tutor: Advisor: Prof. Maurizio Panti Prof. Luca Spalazzi Motivation The network allows any organizations/users to catch up services/information every time everywhere. Producers and consumers are constantly producing and consuming location-dependent and location-independent information. Producers and consumers are considered as autonomous units that need to be coordinated each other as automatically as possible. Generally, at any time, such units have no permanent network connections and no global visibility over partners/units.
Motivation In order to effectively support such scenarios, we must cope with the principal information coordination and integration issues that one expects when dealing with distributed, heterogeneous, and dynamic data: problem reformulation; data integration; source schema updating; consumer s information updating; failure handling. Goals In order to overcome such problems, we must build up an integration and coordination system that owns: an appropriate architecture; efficient query rewriting and data integration algorithms; expressive query language and data description language; an effective cooperation among all the components of the system;
Talk Overview Talk Overview - The Core Agent Platform and the Mediator-based System -- The information source -- The mediator - The Peer Agent System -- The Planner role -- The Strategy components - Conclusions and future work System Architecture System Architecture Software The Core Agent Platform (FIPA 97-98) 98) Mediator CBR-based Information Sources as wrapper agents Agent CBR Mediator Case- Memory Agent Management System Directory Facilitator db1 Wrapper 1 db2 Wrapper 2 Agent Communication Channel (IIOP)
Information Source Information Source A source provides an answer only when a query is equivalent to a portion of the local schema, therefore we need a query rewriting algorithm. Each source adopts a local-as-view (LAV) approach. Source View Representation: Definition. Let s be a source. Let T(s) be a terminology. Let V(s) be a set of concepts defined over T(s). Then S(s) = <T(s),V(s)> Translation Table: is a source view representation. Definition. Let S(s) = <T(s),S(s) = <T(s),V(s)>> be the source view representation of source s. Let V(s) = {V 1,,V n } where V 1,,V n are concepts defined over T(s). Let L be the local query language of the source. Let V 1L,,V nl be queries expressed in L corresponding to V 1,,V n respectively. Then Tt(s) = {<V i,v il > 1 i n} is the translation table of the source s. Information Source Information Source Query reformulation: [Beeri et al.,1997] SRew(Q, V(s)) = V { Q} inf (Q, V(s)) if if V V sup sup (Q, V(s)) φ (Q, V(s)) = φ Q Q Figure: A fragment of the Source View Representation S(w 3 ) of source w 3.
Mediator In our system, a mediated schema is composed by a collection of queries (i.e., views over a mediated schema) and their corresponding reformulation (i.e., views over the sources). Both represented as concepts over C-Classic terminology. Moreover, each mediator can cooperate with sources and other mediators. This allows the mediator to update its mediated schema every time a source schema change or a consumer has a new information need. As a consequence in our system, a mediator adopts a global-local-as-view (GLAV) approach. Mediator GLAV assertion: [Spalazzi et al., 2002] Definition. Let Q be a query. Let S 1,,S n be set of information sources. Let Q 1,,Q n be queries to S 1,,S n respectively. Let Sol(Q) be a set of arbitrary combinations of Q 1,, Q n such that each element of Sol(Q) is subsumed by Q. Then Q is a view over the mediated schema, Sol(Q) are views over the sources, i.e., a rewriting of the query Q, and <Q,Sol(Q),<(Q 1,S 1 ),,(Q n,s n )>> is a GLAV assertion. Mediated View Representation: Definition. Let m be a mediator. Let T(m) be a terminology. Let V(m) be a set of concepts defined over T(m) (views over the mediated schema). Let C(m) be a set of GLAV assertions defined over V(m). Then M(m) = <T(m),V(m),C(m)> is the mediated view representation of mediator m.
Mediator Local query reformulation: [Beeri et al.,1997] { Q} MRew(Q, C(m)) = i ( Xi ) V 1 inf (Q, V (s)) ( Sol( X i i )) if V sup if V sup (Q, V (m)) (Q, V (m)) = Failures: Figure: Taxonomies of failures. Talk Overview Talk Overview - The Core Agent Platform and the Mediator-based System -- The information source -- The mediator - The Peer Agent System -- The Planner role -- The Strategy components - Conclusions and future work
P2P: a definition P2P: a definition This paradigm, popularized by systems such as Napster and Gnutella, views a distributed system as an open, dynamic network of peers (coalition). Peer s main characteristics: each node is equal (peer) to the others and may operate as router, client, and server according to the task to be performed; has an independent addressing system; is able to cope with variable connectivity; uses an instant-messaging communication model. These characteristics make this model quite different from classical computing models (e.g. client-server and Web-based) Multi-Agent Systems Multi-Agent Systems Agent paradigm give us the better design mechanism to model P2P coalitions. Indeed, an agent encapsulates both client and sever capabilities in order to exhibit autonomous and collaborative behavior, i.e., as a peer requires. Moreover, several MASs have been used to support the exchange of services and information among agents into distributed, heterogeneous, and dynamic environments, e.g., the InfoSleuth project [Bayardo et al., 1997] and the system described in [Martìn et al., 1999]. Nevertheless, such works do not stress enough cooperation strategies among agents in order to overcome local failures that especially arise in P2P coalitions.
Peer Agent System Peer Agent System P2P Agent System Architecture by i* Strategic Dependency model [Yu and Mylopoulos, 1995] [Yu and Liu, 2001] users system Agent and Role (Actor). Softgoal dependency Task dependency Resource dependency The strategic Planner role The strategic Planner role When a Peer Agent is not able to satisfy (reformulate) a request (problem) locally this means a failure occurs. The Planner has to build up a cooperation strategy in order to overcome such a failure: distributed problem reformulation. [Penserini, 2002] Reformulation: each time a request/problem arrives the peer agent rewrites the request in terms of its local knowledge, i.e., mediator ability. [Spalazzi et al., 2002] Our analysis focuses on such principal local failure reasons: - inability to rewrite a given problem; - at least one partner, involved in the reformulation, causes the failure; - at least one partner, involved in the reformulation, is not connected; - a proposed solution to a partner is rejected.
Strategy Structure (Plan) Each strategy is a four element structure (plan): <partner, answer, modality, request> where each element contains several alternatives: Chose Partner: All Peers ((PA)), New Peers, Peers Temporarily Disconnected, Active Peers, Peers that succeed, Peers that Failed. Ask for Answer: Ask for Data, Ask for Reformulation, Ask for both Data and Reformulation. Chose Modality: Search for Peers, Wait for Peers, Send Email. Submit Request: Send the original (P), Send the reformulation (R), Send the failed components. Conclusions In this work, we have designed and implemented a multi-agent system prototype based both on the FIPA specification and on a mediator architecture. Available at: http://jeap.inform.unian.it Besides, we have presented a preliminary study on the capabilities of a P2P-based collaborative information system. Moreover, we have proposed a way to model and evaluate cooperation strategies in a P2P agent system. In such a case, i* framework is suitable to figure out and visualize the intentional relationships among and the rational within actors of a P2P coalition.
Future Work Future Work We are improving the functionalities of a MAS (JEAP) in order to cope with the complexity introduced by P2P coalitions. Moreover, we are defining a comprehensive set of scenaria to evaluate our prototype system, and make sure it behaves as expected. In particular, we are working on a case study related to Virtual Enterprise s Supply Chain Management where applying our P2P-based collaborative information system. Some References Some References [Bayardo et al., 1997] Bayardo Jr., R. J., Bohrer, W., Brice, R., Cichocki, A., Fowler, J., Helal, A., Kashyap, V., Ksiezyk, T., Martin, G., Nodine, M., Rashid, M., Rusinkiewicz, M., Shea, R., Unnikrishnan, C., Unruh, A., and Woelk, D.: InfoSleuth: Agent-based semantic integration of information in open and dynamic environments. In Proc. of the ACM SIGMOD international conference on Management of data, Vol. 26, No. 2, June 1997. [Beeri et al., 1997] Beeri, C., Levy, A.Y., and Rousset, M. C.: Rewriting Queries Using Views in Description Logics. In Proc. of the 16th ACM Symposium on Principles of Database Systems (PODS), 99-108, ACM Press, Tucson, Arizona, 1997. [Martín et al., 1999] Martín, F. J., Plaza, E., and Arcos, J. L.: Knowledge and Experience Reuse through Communication among Competent (Peer) Agents. Published in International Journal of Software Engineering and Knowledge Engineering, Vol. 9, No. 3, 319-341, 1999. [Penserini, 2002] Penserini, L.: Integration and Coordination in both Mediator-Based and P2P Systems. Ph.D. Thesis, Istituto di Informatica, University of Ancona, Italy, 2002. [Penserini et al., 2002] Penserini, L., Lin, L., Mylopoulos, J., Panti, M., Spalazzi, L.: Cooperation Strategies for Agent-Based P2P Systems. Accepted in: WIAS: Web Intelligence and Agent Systems: An International Journal, Publisher: IOS Press, ISSN 1570-1263, 2002.
Some References Some References [Spalazzi et al., 2002] Spalazzi, L., Panti, M., Penserini, L.: Evaluation of Cooperation Strategies. TR 2002-11, Computer Science Institute of University of Ancona, Ancona, Italy, 2002. [Yu and Liu, 2001] Yu, E., Liu, L.: Modelling Trust for System Design Using the i* Strategic Actors Frame-work. In: Trust in Cyber-Societies Integrating the Human and Artificial Perspectives. R. Falcone, M. Singh, Y.H. Tan, eds. LNAI-2246, pp.175-194, Springer, 2001. [Yu and Mylopoulos, 1995] Yu, E., Mylopoulos, J.: From E-R to A-R Modelling Strategic Relationships for Business Process Reengineering. Int. Journal of Intelligent and Cooperative Information Systems, 4(2&3), pp.125-144, 1995.