Enabling Selective Flooding to Reduce P2P Traffic

Abstract

We propose a P2P cooperation policy to increase the effectiveness of the flooding-based approach used to retrieve information over pure P2P networks. Flooding consists in propagating the original query from the source peer to “known” peers, and, in turn, to other peers, producing, in the general case, an exponential grow of the search traffic in the network. According to our policy, each peer involved in the flooding process propagates the query only toward peers hopefully capable of satisfying it. The crucial point is: how to detect such good candidates? Of course, “local” properties of similarity between peers not satisfying the transitivity property cannot be used to the above purpose, due to the necessity of propagating queries. Our solution relies on recovering some transitivity behavior in similarity-based P2P information retrieval approaches by considering neighborhood semantic properties. Experimental results show that the selective flooding so obtained is effective in the sense the traffic is drastically reduced w.r.t. the standard flooding (like GNUTELLA), with no loss of query success.