Proactive repair redundancy algorithms for distributed storage in P2P networks

Distributed storage systems frequently use erasure coding based redundancy or replication to improve availability. Traditional erasure coding based approaches to redundancy are optimal in terms of their storage overhead-reliability tradeoff, but are expensive to repair. In this paper, we propose Proactive Repair (PR), a novel hybrid redundancy scheme. We discuss the implementation of the proposed scheme and discuss an emulated-based experimental study to evaluate PR's performance. The results show that PR outperforms exact-repair Minimum Bandwidth Regenerating (exact-MBR) network coding in terms of complexity and repair traffic overhead. Content availability when the system uses PR exhibits higher ability to dynamically adapt to changing failures rates, caused by nodes leaving the network.