In Dynamic NAS (DNAS) backup jobs, getting data quickly off an array snapshot to target backup storage has been a performance challenge. Most array controllers have TCP/IP port limits that throttle throughput. Even worse, snapshot mount hosts didn’t have a graceful way to mount snapshots as a group to speed things up.
The NetBackup 10.4 snapshot crawler has been optimized to deal with these challenges. The parent backup job controls the snapshot crawler’s snapshot crawl for each child job. It breaks up the snapshot into slices of work serviced by a child job for each slice. Each job then runs through a different mount host mounted to the snapshot. In this way multiple mount hosts in a DNAS mount host group can speed up the backup of a single snapshot.
The best part is that child jobs aren’t necessarily limited to running in parallel. After each child job crawls complete and the file/folder worklist is finalized, that child job will run. The child jobs don’t wait for all child jobs to post ready before backup data is read from the snapshot and written to backup storage.
Our testing shows this can greatly increase snapshot backup performance. For example, we saw a 50-hour snapshot backup that used one mount host decrease to a 15-hour job using four mount hosts. This decrease in time only increased array controller port utilization 10-20%. A fair trade indeed.
Users doing disk array snapshot backups should be sure to check out and implement this feature of NetBackup 10.4+.
