摘要

Durable P2P data storage is a research focus in recent years. It aims at providing data for requesters anytime and anywhere. Considering the fact that any peer could join and leave P2P network freely and randomly, how to improve the data availability is a crucial issue when implementing a P2P data storage scheme. To guarantee data availability, data replication and replica management have been found as the effective strategies by distributing the replicas in different nodes. However, how to determine the data redundancy is a key issue which should be solved carefully. In this paper, according to the existing researches, we first calculate peers' online probabilities in each time interval, and then we group the peers which have the complementary online patterns to create the so-called generalized peers with high online probability in each time interval. With the created generalized peers, we present the durable P2P data storage scheme. The simulation results show that our P2P data storage scheme outperforms the existing strategies in terms of the data retrieval success rate and the response time in dynamic P2P networks.