Centralized multi-node repair for minimum storage regenerating codes

摘要

In distributed storage, erasure codes are widely used to provide data reliability, where every codeword symbol corresponds to one storage node. The network traffic cost during the repair of node failures, called repair bandwidth, is an important metric in code design. In particular, minimum storage regenerating (MSR) codes are maximum distance separable (MDS) codes that have optimal repair bandwidth. In this paper, we generalize the problem to minimum storage multi-node regenerating (MSMR) codes, which are MDS codes with optimal repair bandwidth for e node failures. We describe a general framework for converting MSR codes to MSMR codes. The repair strategy for e failures is similar to that for single failure, however certain extra requirements need to be satisfied by the repairing functions for single failure. Then we apply this framework to product-matrix codes and interference alignment codes.