"As computing and communication capabilities have continued to increase, more and more activity is taking place at the edges of the network, typically in homes or on workers desktops. This trend has been demonstrated by the increasing popularity and usability of "peer-to-peer" systems such as Napster and Gnutella. Unfortunately, this popularity has quickly shown the limitations of these systems, particularly in terms of scale. Because the networks form in an ad-hoc manner, they typically make inefficient use of resources. We propose a mechanism, using only local knowledge, to improve the overall performance of peer-to-peer networks based on interests. Peers monitor which other peers frequently respond successfully to their requests for information. When a peer is discovered to frequently provide good results, the peer attempts to move closer to it in the network by creating a new connection with that peer. This leads to clusters of peers with similar interests, and in turn allows us to limit the depth of searches required to find good results. We have implemented our algorithm in the context of a distributed encyclopedia-style information sharing application which is built on top of the gnutella network. In our testing environment, we have shown the ability to greatly reduce the amount of communication resources required to find the desired articles in the encyclopedia."

Finding Good Peers in Peer-to-Peer Networks