A Single Disk Failure Recovery for X-code Based Parallel Storage Systems

Achieving data availability and reliability guarantees against disk failure using redundancy coding scheme. X-code is a double tolerant to achieve the optimal update complexity. It reduces the possibility of data unavailability when disk/node fails. X-code based optimal recovery scheme Minimum-disk-read-recovery (MDRR) and Group-based MDRR (GMDRR) minimizes the number of disk reads for a single-disk failure recovery. A tight lower bound of disk read is formed and MDRR algorithm is applied to match the theoretical lower bound. A disk read cannot be balanced while matching the lower bound of disk reads within a single stripe and cannot be balanced among different disk by simply rotating disk. A leap rotation scheme which balances disk read among different disk within a group of stripes and it matches the lower bound of disk read is called as GMDRR. MDRR reduces around 25% percent of recovery time of the conventional approach.