Hi guys,
I have a question...
My DAG enviroment is composed by three servers every node is a virtual server (VMWARE), but disks where databases (29 databases = 800 GB) are on LUMs.
Netbackup enviroment: Master server is Windows 2012 and media server is 5230 appliance.
Full backups are taking almost 12 hours!! Could anyone give me an idea to improve this!
Sounds to me like the most likely culprit is a storage (i.e. disk read) performance issue on the backup client (i.e. the MS Exchange Servers), because this throughput is so slow:
| 800.0 | GB |
| 819,200.0 | MB |
| 12.0 | hours |
| 43,200.0 | seconds |
| 19.0 | MB/s |
.
How to benchmark the performance of the bpbkar32 process on a Windows client
http://www.veritas.com/docs/000013839
...obviously do not "read" the volumes or folders or files of the MS Exchange databses as part of your test...
...so use some suitable large enough dummy file in a nother folder on the same volume(s)... and see what pure/plain disk read speeds you get...
...if you still see 19MB/s, or thereabouts, then it clearly is a disk storage issue, and not an MS Exchange issue and not a NetBackup issue.
FYI - by the very nature of discreet systems, there will always be a bottleneck of some kind at some point. But I agree 19 MB/s is way way too low for an MS Exchange Server database backup.
You may want to have a good look at Job details and logs.
I have seen at a customer site where the consistency check took more than 2 hours. Per Information store!
Nothing that can be done from NBU side - other than disabling the option for consistency check.
Hi Andres
OK yes, you can't use SAN Transport because of the RDM (in case of VMware type backup) or SAN Client (because its a VM and doesn't have its own FC).
I'd start with reducing the streams and then increasing them again to see where you get the best performance. That will take a few backup runs to figure it out.
I suppose you're running 10G so the network shouldn't be the problem. You can monitor the client's read performance during the backup, and see at what stage it maxes out.
Example
1 stream provides ± 40MBps
2 streams provide ± 80MBps
3 streams provide ± 120MBps
4 streams provide ± 120 MBps <<<<You'll see at some stage that it doesn't help to add more streams. It just loads the processor / memory but there is no incremental improvement in transfer speed.
Any chance to migrate from RDM to VMDK would make your life easier ;)
Hi Riaan,
Across the LAN or the SAN? LAN, actually Im not sure how to configure SAN backups, the DAG´s node are VM´s and there is a policy for those but databases are not inside these VM, databases are on LUMs.
Have you checked the read performance of the disks on the Exchange server?
Do you take full and incrementals, or just fulls? Just fulls because almost everyday I have to perform mailboxes restorations.
Granular or not? yes.
How many streams are you running? 29 streams, one per database.
Across the LAN or the SAN?
Have you checked the read performance of the disks on the Exchange server?
Do you take full and incrementals, or just fulls?
Granular or not?
How many streams are you running?
