Hi,
We have tested using Veeam B&R to backup and recover our SQL2008R2 SAP DB (It has 10 x 60GB Datafiles + log file). Currently we use Backup Exec, backup up to LTO6 tape and it takes 1.5 hrs to backup and a similar time to restore. However the restore with Veeam (from disk) is much, much slower. My belief is that Backup Exec creates all the restore datafiles and populates them in parallel, whereas Veeam creates each 60GB file sequentially, hence the sloweness. Could the developers enhance this so that Veeam can be considered an option for backup and recovery of large Enterprise level SQL DB's? We use Veeam to backup our other 375 VM's, why not remove our last need for Backup Exec.
Does anyone else have a view on this?
Hi Foggy,
The backup is done in SAN mode with a physical server. The restore back to an alternative SQL server VM using 'SQL item restore' mounts the backup VM on the server and then allows the Veeam SQL explorer to redirect the datafiles to the alternate VM. At this point the restore of these datafiles happens sequentially over the network and is dependent on the network speed. We have changed the nic on the physical backup server to a 10Gb and the target VM has a 10Gb vmnet3 adapter, however the maximum throughput we can get is 700mbps. We have testing throughput with other tools and much higher is possible. Also the statistics for a application restore job are rudimentary compared to the level of statistics avaliable for backups - it would be good to see the bottleneck identified. Also it would seem to be possible to restore the multiple datafiles in parrallel given that they are going to different disks, rather than sequentially which is part of the bottleneck. Any advice appreciated.
Restoring multiple datafiles to different disks in parallel seems to be possible, indeed. However, keep in mind that FLR does random read from the storage, so the bottleneck could be there. Btw, what kind of storage it is?