The architecture and performance of topic‐driven P2P resource discovery system

Abstract

SUMMARYMany network applications over peer-to-peer networks, such as Gnutella, have been developed drastically. However, by using distributed systems with simple search functions, it is difficult to discover suitable information resources and to guarantee practical response time. Furthermore, we are faced with several problems of search performance which are caused by the bandwidth of networks and the instability of peer-to-peer network topology. In order to realize good network performance and advanced functions with content-based search, we introduce the index methods with the characteristic of each file. We propose the topic-oriented search mechanism, which selects optimal search hosts, based on heuristic functions with several parameters in each host. We evaluated the performance of our proposed methods on a simulation of a prototype system and confirmed that our system achieves a suitable search performance with capability of responding to changes in network topology.