Decreased VMWare performance due to multiple snapshots?

[Morat]

Hello,
I've got an application vendor who is blaming the poor performance of his product on VMWare - specifically the fact that we back up using Veeam. He is claiming that if you constantly create and delete snapshots then the performance of that VM Host will gradually degrade. He's demanding that we 1. reinstall ESXi and 2. Use a different backup strategy.
Has anyone noticed a gradual performance hit due to using Veeam backup on VMWare? (I'm using ESXi 4.1 on a couple of Dell 2x4 core Poweredge servers with iSCSI SAN).
Thanks!

[Gostev]

the performance of that VM Host will gradually degrade

This is a false statement. There is no way how host performance can be affected by what is happening with the storage (which is where snapshot magic takes place), so this "gradual degradation" claim does not make any sense whatsoever. Basically, it's like saying that using snapshots will gradually reduce host CPU clock frequency or something...

VM performance can in fact be affected for those VMs running highly transactional application generating lots of disk I/O, but only during very short period of time when snapshot files are being committed back into VMDK file (when backup finishes). Because backup is usually scheduled for off-hours, this typically goes unnoticeable. However, if that is still a concern for some VMs running 24/7 apps, there are ways to improve that (such as redirecting snapshot location to the dedicated SSD drive).

[jlehtinen]

I've never seen any issues like that. There can be performance hits while a replica or backup is being pulled, but I've never seen issues with that process causing data corruption in the original VM. If creating snapshots degraded the VM, that would be a big issue for everyone using VMware products...

Maybe what he is talking about is how if you create a snapshot and then don't consolidate it (delete or commit). If you take a snapshot and continue running the VM, the changes to that VM are actually wrote out to a separate file, if the VM continues in that manner for a very long time, or more & more snapshots are taken, delete/commit/restore actions can take a LOOONNNNG time because ESXi has to spin through all those separate files.

So I guess if you had a lot of snapshots in ESXi, PLUS you were running Veeam on top, you might see performance issues. In my setup, I only use snapshots for testing servers, I never use them otherwise for production VM's.

[J1mbo]

Why not clone the VM in question to some other thick-provisioned storage (and isolated network of course) and give that to him to play with?

[Morat]

Thanks for your replies. I have to admit that I was very sceptical about the claims but it's hard to prove a negative! I think the cloned test server is a good idea.
All the best,
Morat