Efficient content location based on interest-cluster in peer-to-peer system

Abstract

Content location is a challenging problem in decentralized peer-to-peer systems. And query-flooding algorithm in Gnutella system suffers from poor scalability and considerable network overhead. Currently, based on the Small-world pattern in the P2P system, a piggyback algorithm called interest-based shortcuts gets a relatively better performance. However, we believe it could be improved and become even more efficient, and a cluster-based algorithm is put forward. The main concern of our algorithm is to narrow the search scope in content location. Resource shortcuts are grouped into clusters according to their contents, and resource queries are only searched in related shortcut clusters, so that the search efficiency is guaranteed and the network bandwidth is saved. In our experiment, cluster-based algorithm uses only 40% shortcuts roughly, compared with the former algorithm, and the same success rate is achieved. At the end, we refer the relationship between cluster-based algorithm and semantic overlay networks, which is a potential kind of overlay in the future