Concurrent Failure Recovery for MSR Regenerating Code via Product Matrix Construction

Abstract

Node failures are very common in distributed storage systems. Regenerating codes can minimize the network bandwidth required to recover the data lost on the failed nodes. Minimum Storage Regenerating (MSR) code is a class of regenerating codes that can maximize the storage efficiency, meanwhile minimizing the repair bandwidth. The original MSR code via Product Matrix (PM) provides a means for single failed node recovery. In this work, an algorithm of recovering multiple failed nodes concurrently with the minimum feasible bandwidth will be proposed, extending the framework of the original PM MSR code. Based on the proposed strategy, the needed bandwidth for centralized and distributed recovery policies, which are the two major categories of repairing policies, will be explicitly expressed against the coding parameters and the number of failed nodes, hence numerical comparison can be made between them. Moreover, the impact of Repairing Degree (the number of surviving nodes from which the assistant data are downloaded ) on the bandwidth cost will be studied to help make optimal decision in practical storage systems.