Expected performance using EMC Data Domain

[ced123]

I have looked at some threads about this, but some are a little old, so apologies for some overlap.

I am just setting up Veeam B&R, using a Data Domain 68000 with DD Boost as primary storage. I believe I have used the recommended settings when creating the repository and backup job (as in KB 1956). For testing purposes, I backed up a single Hyper-V VM with 2 x vhdx drives of about 160GB and 1.5TB respectively.
An initial full backup took just under 4 hours -- which is excellent compared to the 11 hours our current backup system takes. However, restores are a real problem. An attempt to restore the entire VM is taking many hours - after 1.5 hours it has restored only 1% of the total - 4 to 7 MB/s, when the backup ran at about 130.

I am aware that DD is not ideal as primary storage for Veeam and restores, especially FLR will be slow compared to non-deduping storage -- but this slow ???
Quite prepared to believe I may have missed a tick box somewhere -- or am I getting the expected restore throughput?

I'm using on-host proxy for the Hyper-V backup. I have a single-use server (Windows 2016) with plenty of resource as gateway to the DD. 10G network all the way, CPU use on proxy/gateway well under 10%, Memory use on each is well clear of the limit. Veeam 9 3a (updated from 9 3) -- we did some trials earlier this year, but used NetApp storage -- which was significantly faster at restoring.

Thanks
Chris

[Mike Resseler]

Hi Chris,

First: Welcome to the forums!

Did you disable dedupe on our side (in the job settings)?

But even with that said. A lot of these things are environment related so it might be best to create a support ticket so they can dive in the logs and see where the big bottleneck is.

Please post the support case ID here and the outcome after investigation from our engineers

Thanks
Mike

[ced123]

Hi Mike,

Yes, I did disable dedupe on the Veeam side. After further testing it looks as though the issue may be the Hyper-V cluster shared volume, so I'll check this out before creating a support case.

Thanks,
Chris