Adaptive Media Replication in Unstructured P2P File Sharing Systems Based on Geographical Properties and Query Distributions

Abstract

Assuring acceptable delivery performance and resource availability is a very relevant issue in P2P networks. We see that replication strategies in P2P file sharing systems are mainly designed to assure high resource hit rate, rather than to improve average delivery throughput. In this paper we present a new replication strategy in an unstructured and decentralized P2P network that, as well as providing high average hit rate, significantly improves delivery performances in comparison to freenet-based and random replication strategies. Also, we show that the only average request rate is not a good metric to trigger replication processes. So we propose a new locality-aware replication and replacement strategy based on both topological properties and local user preferences. Simulation results show that the proposed strategy provides higher hit rate and average download speed in comparison to Freenet-based and other replication approaches.