Repair Locality with Multiple Erasure Tolerance
In distributed storage systems, using erasure codes instead of straightforward replication may lead to desirable improvements in storage overhead and reliability. A challenge of the coding technique is to efficiently repair the packets loss caused by node failures so that the system keeps the same level of redundancy. However, traditional erasure codes are inefficient in concern with the repair bandwidth as well as the number of disk accesses during the repair process. As an improvement, regenerating codes and codes with repair locality are proposed respectively.