A Scalable Replica Management Method in Peer-to-Peer Distributed Storage Systems.

Large numbers of replicas in peer-to-peer distributed storage systems deteriorate inconsistency and load imbalance. According to those data management problems, a scalable replicas management method based on decentralized and unstructured peer-to-peer network is proposed. Replicas are partitioned into different hierarchies and clusters according to single replica replication, and then replicas are coded and managed based on the user-defined hierarchy-coding rule. After that, replicas are organized with centralization in local and peer to peer in wide area, and the cost of reconciling consistency can be greatly depressed combining with defined propagationtime-plot. The simulation results show it is an effective multi-replica management method, achieving good scalability, and adapting well to applications with frequent updates.