ONTAP Discussions
ONTAP Discussions
Hi all,
We are currently testing out SME on our Exchange 2007 environment using ISCSI. Currently the backup is taking about 1 and a half hours to backup and verify our 12 storage groups (separate luns for each database and lun) including snapmirroring to our other data center. We only have about 100 Mb in each storage group right now (will most likely grow to about 100 Gb when we move our production users), and I'm wondering if this is normal for these sizes of databases.
Thanks in advance.
Ryan
Ryan,
I believe the transaction log LUNs from different storage groups reside on different storage volumes. If you configure the transaction log LUN(s) from different storage groups to reside on same or limited storage volume, the backup performance will definitely improve. You may want to work with the support to correct this issue.
Regards,
-Qing
Hmm.....how long does a manual eseutil run take? (as that is essentially what is going on in the background)
Andrew,
The speed of eseutil verification will depend many factors, but you should see something like:
85MB/sec for iSCSI
185MB/sec for FCP
I have seen verification rates with FCP as high as 256MB/sec.
Factors that can make these rates less:
- Network congestion (for iSCSI)
- Available memory and CPU on verification host
- Number and type of disks in the aggregate where the verification is taking place
- Storage controller utilization
The use of eseutil in an Exchange 2007 environment is deprecated. You should be using CheckSgFiles.dll for database verification.
The CheckSgFiles will give you faster database verification performance.
To enable CheckSgFiles in SME:
1. Open regedit
2. Go to HKLM\Software\Network Appliance\SnapManager for Exchange\Server
3. Set UseCheckSgFilesVerify to 1
4. Close regedit
5. Restart the SnapManager Service
I hope that this helps.
Thanks,
Mark
Thanks for the info guys, but it doesn't seem to be an issue with the verification. We've only got tiny little databases right now, as we are only testing, and the verifications are taking less than 20 seconds per database. It seems that every command that goes to the NetApp is taking a few seconds to respond. Ie. renaming a snapshot takes about 6 seconds according to the log.
Ryan,
How do you have your LUNs laid out? Do you have each log LUN in a dedicated flexvol?
Slow LUN enumeration and backup performance has been observed at other customers.
It is suggested that you put all of your Transaction Log LUNs into a single Flexible Volume.
The reason is that SME iterates through a backup based on the number of Transaction Log Flexible Volumes.
Thanks,
Mark