| 摘要 |
A disk array server has a cache and a log drive wherein data blocks, as received are written synchronously to both the cache and the log drive, the cache being written back to the disk array as opportunity affords. The log drive is managed so, when full, data is overwritten in the order first stored on the log drive. Data blocks written to the log drive are flagged as to whether the same block in cache has been written to the disk array, and the flags are updated as the cache is written back to the disk array. In the event of a power failure, data lost from the volatile cache as not yet written to the disk array may be recovered from the log drive. In one embodiment, the recovery is automatic on startup after a power failure.
|
