Bundling Together RAID Disk Arrays for Greater Protection and Easier Repairs

Large data storage systems often use Reed-Solomon erasure codes to protect their static data against triple or even quadruple disk failures. The main drawback of this solution is the high cost of recovering the contents of failed disks, as it requires reading the contents of all the surviving peers in its parity stripe. We propose to reduce this cost by grouping together a small number of conventional RAID arrays into a single bundle and adding column-wise parity disks such that all disks in a given RAID array belong to a distinct parity stripe. As a result, the number of disks involved in the recovery from a single disk failure will be equal to the number of RAID arrays in the bundle. We show that bundles of RAID level 5 arrays can recover without data loss from all triple and at least 96 percent of quadruple disk failures while bundles of RAID level 6 arrays can similarly recover from all quintuple and 99.9 percent of sextuple disk failures.