A Dynamic Convergent Replication Strategy Based on Distributed Hierarchical Systems

Data replication strategies have been used to increase data access efficiency in P2P systems. This paper studies the replica selection strategy and replication strategy. By concerning real-time bandwidth, performance of the peer, and replica's history request, following the principle that high valued replica has high proportionality to be select, a proportional matching based self-convergent replica selection strategy(PMC) is presented. By concerning the access frequency, the status of the network connection and average response time, and implementing replication only if the logic replica is a popular and scarce one, a hot-estimation based ECO replication strategy (HOT_ECO) is presented in this paper. The strategies avoid leading to high overhead for unnecessary file replications and consistency maintenance in case of updates while replicating. The simulation results show that the optimal selection and replication strategy offers better performance by shortening the response time with fewer replicas.