exchange slow downs

[arsprod]

I've read several threads reporting exchange server performance slowdown during replication during snapshot creation and removal. I'm seeing marked slowdown during the actual replication - user performance is really effected (and we're a pretty small shop). Before I add resources to the exchange vm I want to see if others have experienced the same problem and potential fixes. Thanks,

Aaron

[Gostev]

Hi Aaron, this can only happen if your snapshot file location is placed on slow storage. By default, in vSphere 4, snapshots for all VM disks are created in the same folder where VMX resides. vSphere 5 changes this behavior to create snapshot right next to the disk file though. Thanks!

[arsprod]

Hmmm... so if my exchange box is running on my primary san you're saying the snapshot is created there also (which makes sense) and that my slowdown issue is likely the san?

[Gostev]

No, I am saying that snapshot location depends on vSphere version and VM configuration (I cannot know if this is your primary SAN or some other storage that is to blame here, because I know nothing about your environment and Exchange VM configuration). However, yes - the only reason for performance slowdown while VM is running off snapshot is slow storage where snapshot disks are created.

[arsprod]

Sorry to be dense, but I want to be clear before digging into san issues. I'm running esxi 4.1, exchange box is running on one lun on primary san, replicas are saved to windows 2008r2 box running starwind iscsi. If I understand correctly, the snapshot files are being saved to the primary san. I am getting this correctly?

[Gostev]

Correct. And if you have multiple virtual disks for your Exchange, all snapshot files will be created on the same LUN (where VMX resides).

[arsprod]

thanks!

[averylarry]

I think it's a bit simplistic to say it can only be a storage issue. (Sorry Gostev.) It could be related to running the replication job, instead of being related to running the VM from a snapshot . . .

It could be high network traffic from the replication job (slamming a 100 Mb network switch, for instance).
It could be high CPU utilization cutting into the Exchange server's resources.
It could be high memory usage cutting into the Exchange server's resources (possibly generating excessive pagefile usage).


You can test this by creating a snapshot directly without running a replication job and see if you have performance problems.

[joergr]

Most people don´t believe this, but in my experience, more than 90% off ALL (yeah that´s right, i mean ALL) vSphere related Support-Cases DO IN FACT rely on storage performance or storage misconfiguration. BTW, thats the way VMware sees it, also

My personal opinion: Storage is THE main and most important thing in ANY vSphere environment.

Best regards,
Joerg