Optimal Repair Algorithm of Single-Disk Failure for Array Codes with Local Properties

Distributed storage systems ensure data reliability and availability in the event of a disk failure by employing array codes. Generalized Expanded Blaum-Roth (GEBR) codes and Generalized Expanded Independent Parity(GEIP) codes are MDS array codes with local properties that can tolerate multiple disk failures while maintaining good repair performance. However, when repairing a single-disk failure, the repair bandwidth (defined as the total number of symbols downloaded from other surviving disks in repairing the single-disk failure) is not optimal. To address this issue, we propose a random search algorithm that uses a simulated annealing algorithm to optimize the repair strategy. It starts with a viable repair strategy and gradually replaces the current strategy with another that downloads less data. Experimental results demonstrate that our proposed algorithm can significantly reduce the data required to repair single-disk failure for both GEBR and GEIP codes by 25%-38%, under the selection of suitable parameters, when compared with traditional repair methods. Moreover, our repair algorithm can be done in polynomial time.