Why does Veritas recommend disabling RAID caching on the disk where the deduplication disk storage device is located?
The Backup Exec Deduplication folder depends on writes making it to the disk. As with a database, once data is committed to the disk, Backup Exec needs to know that the data is guaranteed to be written to disk. Turning off write caching ensures that the write does not return until the write made it all the way to the disk.
Enabling write caching speeds up write activity to the disk, and speeds up backups. This happens because Backup Exec doesn't have to wait for the write to finish before continuing on to get the next bit of data to write. However, this happens at the expense of knowing that the write made it all the way to the disk. If the system crashes for any reason ( e.g. power failure, hardware failure, etc), data that is in the write cache may not make it to disk and can result in catastrophic loss of the deduplication folder.
Some RAID configurations deal with this problem by adding battery backup to the disk subsystem to ensure that data in the cache is guaranteed to make it out to the disk. This allows the user to get the speed benefits of having a write cache and the security of knowing that data will make it out to disk even in the event of a power failure.
Enabling write caching on a RAID system is a decision that must be made by the user of the system. If the RAID can effectively handle writing out the cache even in the face of a system failure, then enabling the write cache is an option. However, the user must be aware that if the failure occurs during a critical write and the RAID subsystem fails to properly write out the cache, the deduplication folder's consistency can be compromised which then results in the inability to bring the deduplication folder online for backup and restore operations.