Content-based search in peer-to-peer networks

a well-known P2P system, uses resources inefficiently when directly applied to information retrieval problems. In this paper we propose an efficient search mechanism that extends the standard Gnutella protocol to support content-based retrieval in P2P networks. The idea is to estimate locally the relevance of peers when they receive query messages. Only those peers esti- mated as relevant will retrieve the query and send response messages back to the source. Based on a large real testbed evaluation, we show that our method improves the tradeoff be- tween the quality of retrieval and resources consumed while preserving most advantages of standard Gnutella.