Hi.
We've been using regular streaming technique to do backups from the active node in our Exchange 2007 CCR cluster. The typical throughput that we've seen over the years has been around 50-60 MB per second. Now, we've changed the backup type to VSS based when backing up 15 out of 48 databases on the Exchange server and the throughput for the VSS based job is 30-40MB per second. The remaining databases are still being backed up using streaming backups and they remain at a 50-60 MB throughput.
What I'm looking for now is any hints on how to tweak this, or if what I'm experiencing is expected behavior. If it is indeed expected behavior I'd appreciate it if someone could explain what might case this drop in performance.
Currently, the client snapshot options are set to 1-0-1 and on the CCR machines I'm currently letting Windows decide where it should place the snapshot.
* Will my backup throughput improve if I make sure that the snapshot is always placed on a different volume than the one that is being backed up?
* Doing backups during business hours will naturally make my vss snapshot bigger since there will be more changes to the disk if I understand snapshots and copy-on-write correctly but will this affect the backup performance?
* It seemed as if a VSS backup of databases on a low-utilized volume had a higher throughput but after migrating more users and data onto the volume the performance started to drop - Can this be true or am I mistaking?
* Currently we're backing directly to tape but if this slow performance continues it will be an issue because the tape drives will have to start and stop all the time instead of streaming. I guess that this will also shorten the lifetime of the tapes, besides giving a slow overall performance?
Grateful for any suggestions or hints.
/Tobbe