Abstract

A disk drive within a disk array is utilized to capture the original image of data blocks that are updated, i.e., written over, through normal array processes during backup operations. The method captures original data images in a manner that allows the array to be restored to the state that existed at the initiation of the backup process. During execution of backup procedures data is moved in logical block sequence (0 to N) from the array to a backup device, such as a magnetic tape backup device, continuing until all array data has been transferred. Should a write request be received by the disk array controller during backup, the block address associated with the write request is checked to determine if the original data at that address has been written to the backup device. If the original data residing at the target block address has been written to the backup device the write request is scheduled for execution by the array controller. However, if the original data residing at the target block address has not yet been written to the backup device, the original images of the data blocks are copied to the spare drive prior to scheduled execution of the write request. Upon completion of the backup process, the original data images copied to the spare drive are also transferred to the backup device. Thus, all information necessary to restore the array to the state that existed at the time of initiation of the backup process is saved to the backup device. Although there are other sequences in which data could be moved to the backup device, the particular sequence described herein has the least impact (lowest overhead) on normal array activity during the backup process.