Analysis of Erasure Coding in a Peer to Peer Backup System

Abstract — The peer to peer architecture model is widely used today for distributed systems deployment, which is increasingly becoming used as a model for disk back up systems. Currently, popular peer to peer back up systems use replication to ensure data redundancy, imposing a significant burden on system resources such as storage. Could the technique of erasure coding constitute a more suitable data redundancy scheme for such systems? In this paper we explore the prospect of deploying erasure coding in a peer to peer back up system and analyze how it performs against replication. Erasure coding and replication are compared in terms of CPU and network utilization, storage overhead, fault tolerance, end to end delay, and scalability. In addition to the performance analysis we also determine the turning point between erasure coding and replication with regard to peer availability. We provide suggestions for erasure coding parameters with respect to peer availability and file size based on this analysis. I