Peer-to-Peer (P2P) applications such as Napster and Gnutella have made the Internet a popular medium for re-source and information exchange between thousands of participating users. A primary consideration in the design of such applications is the high network traffic that they generate when searching for resources/information. One can argue that with infinite storage capacity it is possible to maintain complete auxiliary information for all the
resources of a P2P network at a peer node, leading to extremely efficient searches. However, this not only incurs high storage overheads, but also additional costs for updating the auxiliary information when nodes join/leave the system or when data content is updated at a peer node, which can happen quite frequently in a dynamic P2P sys-tem. This trade-off between storage space vs. network traffic opens up a rich research space to explore. Previous research has looked into one possible mechanism, local index, in this space. In this paper, we propose three new mechanisms based on signature files within this space that can provide a better focused search at a lower storage overhead than local index. We have shown that, with a very small storage cost, signatures are quite effective at reducing search costs compared to local index. In addition, the message overheads of join/leave/update opera-tions are adequately compensated by the savings in search messages. Of the three schemes, CN, PN-S and PN-A, that we propose, PN-A gives the best performance.
The schemes have been extensively evaluated through analytical experiments and simulation experiments both with the intention of fine-tuning the parameters that they use (neighborhood radius, storage size) and comparing with the previous proposals using both flooding and single-path search strategies under different network topolo-gies (uniform and power-law), different sizes of key attribute, different number of data items at a peer, different data distribution patterns, different degrees of data replication, and different proportions of operations (search/update). We uniformly find PN-A gives good savings in message volume over Gnutella, random walk and local index approaches at a small storage cost. In addition to the performance and storage savings with signa-tures, there are a couple of other advantages that they exhibit compared to index-based approaches: (a) Signature approaches can search across multiple attributes by appropriately encoding all the attributes when composing a signature, instead of being restricted to one or a small number of attributes which needs to be predetermined as in index approach. This facilitates keyword and content based search. (b) It takes a certain minimum amount (threshold) of storage to store an index. With storage size less than this threshold, index approach can not be used and we have to resort to broadcasts/flooding. On the other hand, signatures do not impose any such restrictions and can work with any amount of space allotted to them (though when the space gets too small the ability to fo-cus the search diminishes). All these observations lead us to believe that PN-A is an extremely popular mecha-nism for implementing resource and information lookup operations in P2P networks.
Our ongoing work is looking into reducing false positive effects in signatures by exploiting real data patterns.
While we have demonstrated in [13] that semantic clustering can improve performance significantly in structured P2P overlays, we are also investigating semantic clustering via signature approaches in unstructured overlays.
Finally, we are investigating P2P applications overlaid on wireless networks.
References
[1] Napster website. http://www.napster.com.
[2] Gnutella website. http://gnutella.wego.com.
[3] D. Kossmann, “The state of the art in distributed query processing,” ACM Computer Survey, Vol.32, No.4, pp.422–469, 2000.
[4] L. A. Adamic, R. M. Lukose, A. R. Puniyani, and B. A. Huberman, “Search in power-law networks,” Physics Review E, Vol.64, pp.46135–46143, 2001.
[5] Q. Lv, P. Cao, E. Cohen, K. Li, and S. Shenker, “Search and replication in unstructured peer-to-peer networks,” Proceed-ings of ACM International Conference on Supercomputing, pp. 84–95, June 2002.
[6] S. Ratnasamy, P. Francis, M. Handley, R. M. Karp, and S. Schenker, “A scalable content-addressable network,” Proceed-ings of ACM SIGCOMM, pp. 161–172, August 2001.
[7] I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan, “Chord: A scalable peer-to-peer lookup service for Internet applications,” Proceedings of ACM SIGCOMM, pp. 149–160, August 2001.
[8] A. I. T. Rowstron and P. Druschel, “Pastry: Scalable, distributed object location and routing for large-scale peer-to-peer systems,” Proceedings of IFIP/ACM International Conference on Distributed Systems Platforms (Middleware), pp. 329–
350, November 2001.
[9] B. Y. Zhao, L. Huang, J. Stribling, S. C. Rhea, A. D. Joseph, and J. D. Kubiatowicz, “Tapestry: A Resilient Global-scale Overlay for Service Deployment,” IEEE Journal on Selected Areas in Communications, Vol. 22, No. 1, pp. 41-53, January 2004.
[10] M. Li, W.-C. Lee, and A. Sivasubramaniam, “Semantic Small World: An overlay network for peer-to-peer search,”
Proceedings of International Conference on Network Protocols (ICNP), pp. 228–238, October 2004.
[11] Morpheus website. http://www.musiccity.com.
[12] A. Crespo and H. Garcia-Molina, “Routing indices for peer-to-peer systems,” Proceedings of the 22nd IEEE Interna-tional Conference on Distributed Computing Systems (ICDCS), pp. 23–34, July 2002.
[13] J. Kubiatowicz et al., “Oceanstore: An architecture for global-scale persistent storage,” Proceedings of the 9th Interna-tional Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), pp. 190–
201, Novermber 2000.
[14] B. Yang and H. Garcia-Molina, “Improving search in peer-to-peer networks,” Proceedings of International Conference on Distributed Computing Systems (ICDCS), pp. 5–14, July 2002.
[15] M. Li, W.-C. Lee, and A. Sivasubramaniam, “Neighborhood signatures for searching P2P networks,” Proceedings of International Database Engineering and Application Symposium (IDEAS), pp. 149–158, July 2003.
[16] Freenet website. http://www.freenet.com.
[17] F. M. Cuenca-Acuna, C. Peery, R. P. Martin, and T. D. Nguyen, “PlanetP: using gossiping and random replication to support reliable peer-to-peer content search and retrieval,” Proceedings of the 12th International Symposium on High Performance Distributed Computing (HPDC), pp. 236–249, June 2003.
[18] C. Faloutsos and S. Christodoulakis, “Signature files: An access method for documents and its analytical performance evaluation,” ACM Transaction on Office Information Systems, Vol.2, No.4, pp.267–288, October 1984.
[19] S. Staissny, “Mathematical analysis of various superimposed coding methods,” American Documentation, Vol.11, No.2, pp.155– 169, 1960.
[20] M. A. Jovanovic, F. S. Annexstein, and K. A. Berman, “Modeling peer-to-peer network topologies through ‘small world’ models and power laws,” Proceedings of Telecommunications Forum (TELFOR), November 2001.
[21] B. Yang and H. Garcia-Molina, “Designing a super-peer network,” Proceedings of International Conference on Data Engineering (ICDE), pp.49–62, March 2003.