This article can also be found in the Premium Editorial Download "Storage magazine: Lessons learned from creating and managing a scalable SAN."
Download it now to read this article plus other related content.
As vendors integrate disk into the backup process and claim support for features like disk-to-disk-to-tape (D2D2T), users need to verify how vendors are implementing this functionality. One of the least desirable ways to implement D2D2T is when the backup software treats the disk as a target and dumps the data to disk. This requires the backup software to back up the dump file again from disk to tape, and two restores must be done: first from tape back to disk, and then from disk to the original location. It can also result in the backup software losing track of the dump files and the contents of the files.
|Backup software: Core features|
|Click here for a comprehensive list of backup software: Core features (PDF).|
EMC employs a variation of this approach to centrally manage backup files created by its SMB Retrospect product. Retrospect will back up data on remote-office servers and desktops to disk located in the remote office. Central data centers can then use EMC's enterprise NetWorker to gather the dump files created by Retrospect from the remote offices and store them centrally. However, multiple administrators may need to get involved to restore files because Retrospect may not know where the dump file resides if the remote office's disk copy has been deleted.
A better approach is to spool the data to disk before moving it to tape. Spooled files are temporarily stored in cache memory until moved to tape. The backup software tracks the files the entire time. (When the files are dumped, the backup software assumes the files will be stored on disk permanently; if they're moved, the backup software won't know where the files now reside.) Using disk as a staging area speeds up the backup, but may slow the recovery; if the backup software has already moved the data to tape, users need to recover the data from tape.
Vendors still employ disk caches to satisfy users who execute multiple backup jobs at the same time and send the data directly to a single tape drive, requiring the backup software to multiplex or interleave backup jobs to put all of the data on tape. While this speeds up the backup job and allows the job to complete within the backup window, it also makes recoveries slower since the data is fragmented; it may also cause backhitches, a stop and go of the tape drive, during the recovery.
Atempo's Time Navigator has the ability to defragment large data blocks. It first creates virtual buckets on disk where data from multiple backup jobs is sent. Data from each of the multiple running backup jobs is then separated by individual backup job and put on tape. This eliminates the need to interleave the data on the tape during the backup and improves recovery performance.
This was first published in July 2006