Research On Replication Management Related Problems In Wide-Area Storage System Based On P2P

Growth in the volume of data promotes that computer storage technology continues to develop. Comparing with traditional distributed systems, P2P technology has unparalleled advantages and broad application prospects, which make it becoming the most popular technology in 21st century.As an important component of the P2P system, data replication is a well-known and widely accepted technique to reduce data response time and network bandwidth consuming. However, it also brings a series of replica management problems that affect the performance of the system.The replica management problems consist of replica creation, replica locating, replica selection and consistency management.This paper studies based on the background of "Based Middleware Platform for the Development and Production" project, and the specific background platform is the WAN P2P storage system based on chord. To provide a load-balanced high-performance, and high-scalable replica management, the thesis explores: replica creation strategy and replica consistency maintain model.Replicas creation strategies tend to make decisions on where and when to create replicas. In this paper, we propose our replicas creating strategy based on investigating and analyzing the related researches. The strategy is proposed by considering the location of the client and the profit. In the replica create strategy driven by profit, each node in the system just consider if the creation can bring profit to it self, which will lead to a load balanced system.Considering the characteristic of P2P system based on chord, the thesis proposes a replica consistency management model based on replica tree. This model is proposed based on the way the chord ring organizes and restructures, and so save the cost on maintaining the structure of the tree. And the thesis also optimizes the gossip algorithm according to users' access history, for each node to maintain a "friendly" neighbor set.In order to evaluate the mechanisms above, simulation is adopted. The experimental results show that our replica management mechanisms are correct and effective. The results is a valuable reference to the application of the mechanisms.