A Reliable Scheme for Cluster Storage System

An important issue in the design of large-scale cluster storage systems is their vulnerability to nodes crash and network failures. To explore the feasibility of implementing a reliable, high performance distributed cluster storage system, we adopt a reliable scheme that data are distributed across storage nodes using erasure coding with small low-density parity-check (LDPC) code. We implement our new scheme in RSC, a proof-of-concept reliable cluster storage system, along with two other well known RAID-1 and RAID-5 schemes. Performance test results show that such a reliable scheme can provide high reliability while keeping the performance overhead small.