Enabling Technologies for Wireless E-Business phần 7 ppsx

229

Case 3: Distributed Market Places

Test Data

Number of physical nodes (machines): 8

Number of seller agents: 10

Number of distributed market places: 4

Comments. Refer to Fig. 9.22. When the number of buyer agents is not large

(less than 50), the application performance of tuple space is not worse than that of

ACL. As we already know that, in message passing, agents in one container send

their messages sequentially. In the case of using distributed tuple spaces, the

agents in a container can access the different tuple spaces concurrently. This pro￾vides real concurrent activities of agent couplings. When the number of buyer

agents exceeds 50, the performance of tuple space becomes worse because more

synchronized access to the distributed tuple spaces block the accessing processes.

Distributed market place seems to result in a lower breakpoint (from 70 to 50

compared with centralized market place). This may be attributed to the behavioral

pattern of agents in the use of these markets and the actual concurrency that is

achieved.

Case 4:Group Purchase

Test Data

Number of physical nodes (machines): 8

Number of seller agents: 10

Number of distributed market places: 4

Total number of buyer agents: 30

Comments. Refer to Fig. 9.23. Tuple space supports agent collaborations more

efficiently than message passing in all cases of group buying. When the group size

increases, more buyer agents join the group in buying. Once the group relationship

is established, communication loads between agents decrease as a much less num￾ber of buying transactions are involved. As a result, the performance gets better.

Moreover, the characteristics of tuple space (global information sharing, bulk

template match, and reactive tuples) introduced in TSAF seem to have a strong

positive impact on agent collaborations.

From this comparison, though not very exhaustive, we observe that using

tuple spaces as the coupling medium does not produce worse application

performance. In cases where distributed tuple spaces exist and agents require

tighter coordination as in group buying, tuple space has a performance advan￾tage over message passing. However, when heavily synchronized communication

is involved, there is a breakpoint, beyond which the load increase induces

sharp decline in performance due to the associative semantics of the tuple space.

This should be viewed in light of the fact that application system development

is relatively simpler as TSAF supports features that naturally occur in applications

like e-business. The TSAF framework will be made available for researchers

from a Concordia Website soon.

9 Multiagent Communication for e-Business Using Tuple Spaces

230

Fig. 9.23. Curves of performance in a group purchase

9.6 Summary

The use of the agent technology in building e-business applications seems to have

several advantages. In this effort coordination among multiple agents is the central

issue. In this chapter we have promoted the role that “tuple spaces” can play in the

coordination and information sharing among multiple agents. Dynamic relations

among the roles played by the agents and proactive as well as reactive responses

can be well supported in a natural way when tuple spaces are used. The advan￾tages of simplicity and flexibility provided by such an approach, we believe,

would outweigh the extra overhead that might occur in certain circumstances as

shown in the case study.

H.F. Li et al.

231

References

1. N. Carriero, D. Gelernter. Linda in context. Communication of the ACM,

vol. 32, no. 4, pp. 444–458, April 1989

2. E.A. Kendall. Role model designs and implementations with aspect-oriented

programming. In Proceedings of ACM Conference on Object-Oriented Sys￾tems, Languages, and Applications, Denver, Colorado, United States, 1999,

pp. 353–369

3. FIPA-1. Publicly Available Implementations of FIPA Specifications:

http://www.fipa.org/

4. D. Gelernter. Generative communication in Linda. ACM Transactions on

Programming Languages and Systems (TOPLAS), vol. 7, pp. 80–112, 1985

5. G. Girard, H.F. Li. Evaluation of two optimized protocols for sequential

consistency. In Proceedings of the 32nd Annual Hawaii International Con￾ference on System Sciences, January 1999

6. G. Cabri, L. Leonardi, and F. Zambonelli. MARS: A programmable coordi￾nation architecture for mobile agents. Internet Computing, vol. 4, no. 4,

pp. 26–35, 2000

7. Rowstron. WCL: A co-ordination language for geographically distributed

agents. World Wide Web Journal, vol. 1, no. 3, 1998, pp. 167–179

8. JavaSpaces: http://www.cdegroot.com/cgi-bin/jini/JavaSpace JADE, http://

sharon.cselt.it/projects/jade/

9. J. Snyder, R. Menezes. Using logical operators as an extended coordination

mechanism in Linda. In Proceedings of Coordination 2002, York, England,

April 2002

10. FIPA-2. Interaction Protocol Specifications: http://www.fipa.org/repository/

ips.php3

11. Y. Li. Reactive tuple space for a mobile agent platform. Master Thesis,

Dept. of Computer Science, Concordia University, Montreal, 2004

12. G. Rimassa. Runtime support for distributed multi-agent systems, Ph.D.

Thesis, University of Parma, January 2003

13. Y. Zhang. A tuple space based agent programming framework. Master Thesis,

Dept. of Computer Science, Concordia University, Montreal, 2004

14. H.A. Mallot. Behavior-oriented approaches to cognition: Theoretical per￾spectives, Theory in Biosciences. Vol. 116, 1997, pp. 196–220

15. G. Wagner. A Logical and Operational Model of Scalable Knowledge-and

Perception-Based Agent. In Proceedings Seventh European Workshop on

Modeling Autonomous Agents in a Multi-Agent World: Agents Breaking

Away, Einhoven, The Netherlands, 1996, pp. 26–41

16. G. Cabri. Role-based infrastructures for agents. Eighth IEEE Workshop on

Future Trends Distributed Computing System, Bologna, Italy, 31 October–2

November 2001

17. T. Dierks, C. Allen (1999) The TLS protocol version 1.0. http://www.

ietf.org/rfc/rfc2246.txt

18. J. Schiller (2000) Mobile Communications. Addison-Wesley, New York.

9 Multiagent Communication for e-Business Using Tuple Spaces