Cooperative repair with minimum-storage regenerating codes for distributed storage
Jun Li, Baochun Li · 2014
Distributed storage systems store redundant data to tolerate failures of storage nodes and lost data should be repaired when storage nodes fail. A class of MDS codes, called minimum-storage regenerating (MSR) codes, has been designed to optimize bandwidth consumption when repairing one single failure. Compared with repairing failures individually, the cooperative repair of multiple failures can help to further save bandwidth consumption when multiple failures are being repaired. In this paper, we present a new construction of minimum-storage cooperative regenerating (MSCR) codes that repair two failures cooperatively and exactly. We show that given a valid instance of linear exact MSR codes, we are able to construct a corresponding repair procedure to repair any two failures cooperatively with optimal bandwidth consumption, i.e., to construct an instance of exact MSCR codes directly from exact MSR codes. With this connection, we are also able to repair any single failure exactly with MSCR codes.