Time-Related Replication for P2P Storage System

Many P2P based wide-area storage systems are appeared to provide scalable storage services with idle resource from many unreliable clients. To minimize the data loss, quick replication is important to replace lost redundancy on other nodes in reaction to failures. The popular approach is the availability based replication which uses the individual node availability. However, with the high churn some replicas leave or fail within similar time. As a result, it requires bursty data traffic and sometimes it loses data. This paper explores the time-related replication which uses the information of the session time to prevent the bursty failures. It tries to get the primary replica which has enough time to replace the lost redundancy. Moreover, the sparse replicas spread on the timeline and the number of overlapped replicas lessens as best it can. Results from a simulation study suggest that the time-related replication keep the high data availability with less data copy traffic.