Replication Strategy in Unstructured Peer-to-Peer Systems

摘要

The unstructured peer-to-peer (P2P) systems usually use a "blind search" method to find the requested data object by propagating a query to a number of peers randomly. In order to increase the success rate of blind search, replication techniques are widely used in these systems. Most P2P systems replicate the most frequently accessed data objects to improve system performance. However, existing replication strategies cannot answer the question that how many replicas of an object should be kept in the P2P system. If an object is replicated excessively, it inevitably will affect the average efficiency of a replica, which will decrease the whole search performance. This paper addresses the issue of finding the proper number of replicas for an object according to its query rate. In this paper, we firstly investigate the precise relation among success rate, the allocation of replicas and query rate. Then we propose an approach of the allocation of copies to optimize the success rate. As a benchmark, our result offers a new understanding of replication.