| 摘要 |
A hot spare drive within a disk array mirrors writes to a selected drive within the disk array. The mirror is light weight in that little effect on system performance is seen, since both writes need not be complete before the write is returned as successful. Upon subsequent failure of a drive within the array, the need for data regeneration is reduced or eliminated if the failed drive is the mirrored drive. The hot spare remains available for failure recovery when a different drive (other than the mirrored drive) fails, in which case conventional data regeneration techniques may be employed. When the mirrored drive fails, however, the period during which the disk array operates with diminished performance is reduced since data from the failed drive is available on the hot spare and regeneration is only required for data not completely mirrored. Even when a drive is arbitrarily selected within the array for mirroring, an overall reduction in the time required for data regeneration of failed disks will be seen in systems employing large numbers of arrays. Error information relating to the disk drives within the drive array and/or predictive failure analysis may be employed to increase the probability that the failed drive was mirrored. The prospect of complete data loss due to multiple drive failure is also reduced, since multiple drive failure is not fatal if one of the failed drives was mirrored.
|
