Veeam Replicas of Exchange 2007

This is the nature of processing.
When the job starts the guest is put into snapshot, this process is therefore going to quiesce the disk as the snapshot is established. During this process your server is going to pause momentarily.
When the guest is taken out of snapshot it will perform the reverse of this process and again pause the server for a new milliseconds.
If this is taking longer than is acceptable there are a few things to look at.
How busy is your storage, more IO means a longer pause.
Are your virtual disks aligned (have a look at vOptimiser) Missaligned disk would be doubling your IO.
I Assume your connecting to your storage via iSCSI. how busy are the NIC's and network.
Is your host server over utilised?
Is your guest server over utilised?

Try and re-schedule the backup job, to a quieter time, so clients do not detect the Snapshot generated by Veeam Backup.

Are the Snapshots being created and deleted correctly?

We run replicas every hour for DR purposes.  We run a backup at night.
I have noticed that at least once a week when I run the "VSSADMIN list writers" command the Microsoft Exchange Writer is in a failed state.  I have to reboot Exchange to corect this or at times it will correct itself.  The server is on the latest SP for Windows and Exchange.
The snapshots are being created and deleted correctly.

During the replicas neither the Exchange Server or the server running Veeam show high performance levels or processes using too much CPU.

I will look into vOptimiser, I have never used it.  

Don't jump to conclusions here with disk alignment.

This can be normal behaviour, for BUSY servers.

Can you check the following, when a backup/snapshot is being created or deleted, can you ping the server and see if you any any timeouts.

we worked with many EE folk, that are having similiar issues, with Snapshotting, causing "gaps" in service for a few seconds.

Look at this solution (I warn you it's long)

https://www.experts-exchange.com/questions/26731168/slow-snapshot-removal.html

But the end result, was a SLOW datastore resulting in slow snapshots. This problem was quite severe, but similar.

Sorry for the delay, just getting back to this.  I looked at the solution and I think my data stores are in good condition. They are on a SAN with 10K SAS drives in RAID5 configuration.  

I did a manual snap shot of Exchange and some users reported their Outlook locked up and others had no problems at all.  I ran a ping to the server during the snapshot and it never timed out but the responce times went from 4000ms at the begining to between 20 and 100ms during the snapshot. Normally they are less than than 1ms.

hanccocka:
Disk alignment. It appears you don't appreciate the lack of attention that is placed on alignment, and therefore the often serious implications.
Virtually all Windows servers pre windows 2008 have alihnment issues.
This will without question double your IO, of course if you don't have a stressed environment the alignment issue will have no effect.
I expect at some time or another all environments experience heavy resource utilisation.
I know our environment is heavily utilised every evening as backups and other maintanence tasks are run. When we had 300 misaligned Windows Servers the IO was unpleasant to say the least.
Again I say don't underestimate the importance of aligning your disks.  

MrLandShark:
Are you using the same NIC's for your Guest connectivity and Storage IP traffic? how do you connect to your storage and at what speed.
Putting a guest into snapshot in a healthy and not over utiliised environment should not cause lockups or timeouts.
But to sort this out you are likely going to have to commit significant time to it. If you don't sort it out it will only get worse as you environment gets busier.

@markzz: We know all about disk alignment my friend!!!  (we've also published on the issues as well!)

What I stated was don't jumpt to conclusions about it, you didn''t even ask What OS he was running, to assume disk alignment was the issue. (if disks are misaligned, it's a huge job to re-align the partition)

MOST, VMware Virtual Machines are NOT aligned, that we have examined. It depends on the environment as to whether it causes performance issues.

Do you any hard stats of before and after correctly aligned disks? That you can publish. (we do!)

Long ping times are to be expected when the Snaphost is taken, when the server is busy. Sometimes you could even expect a "request timeout".  Is this on Create of Snapshot or Write?

RAID 5, 10k is not the best configuration for "high performance based SAN disks". Do you have any other VMs on the LUN?

I assume you are using the Software iSCSI Initiator?

Are you also using jumbo frames, we've seen 100x performance increase when using jumbo frames with iSCSI software initiator, which was better than hardware iSCSI initiators.

The long ping times were on the creation of the snap shot.
There are other VMs on the LUN.
We are not using Jumbo Frames.

On a related issue, our Veeam Replicas and Backups of our Exchange Server have been failing the last three days.  The error messages says "another task already in progress".  There are no other tasks in progress. I have rebooted the Exchange and Veeam servers twice and I still get the error.  How do I convince Veeam there are no tasks in progress?

I was able to get the veeam jobs running again by turning off the Exchange server for a few minutes.  When I turned it back on whatever the task was that was in progress had stopped so the jobs were able to start and finish.

We are still having the problem of Outlook performance issues when the job starts or stops.  Do you know of any products that can replicate Exchange without performance issues?

I have a VM consultant coming out to help re-organize our VM structure. Update ESX, use jumbo frames, realign the data stores...

I would like tom award the points to hanccocka, he was helpful in working on this problem.

Tips helpped understand the problem better but problem still existz.  There appears to be no good answer.