Veeam Restore to Azure - Optimising Performance?

[publicemailnotallowe]

Hi,

I'd like to know anyone's experiences with making the Veeam Restore to Azure functionality as performant as possible on the Azure end. I'm thinking of things like the size of the Veeam Azure Proxy VM, the target Storage Account, the VM-size for the VM that's being restored etc.

Currently I'm hitting a bottleneck somewhere as the restore rate (MB/s) reported during the job is far below the total network bandwidth available.

So far I've experimented with increasing the size of the proxy VM, converted it to use managed disks (SSDs), found that using a Premium Storage account causes a job failure error (couldn't deploy a proxy VM), and choosing a large size for the VM that's being restored. I also created a VM in Azure, installed Veeam B&R, copied up and imported a backup then ran the restore to Azure, but still found a slow restore rate.

If anyone can share their own experiences I'd be really grateful for any advice you can provide.

Thanks

[veremin]

Can you provide us with a bit more details regaring your setup:

- where your backup repository is located?
- what is used as backup repository? if azure vm, what size and what disk type does it have?
- where your backup server is located?
- are you using Azure Proxy? If so, what size does it have?
- what type of connection you have between between repository, Azure Proxy and Azure Infrastructure?
- what is identified as a bottleneck in the session statistics?
- what is the resulting speed?
- etc.

Thanks!

[publicemailnotallowe]

Hi Veremin,

Thanks for any advice you can provide.

Unfortunately the Restore to Azure job details don't provide bottleneck information/stats you get in other parts of Veeam B&R.

I've tried restores from both the on-prem backup server and repo (a SOBR) and from a temporary backup server I built in Azure, which used the local disk as a temporary repo. And yes a Veeam Azure Proxy has been used for all restores as this is recommended. In all tests the restore rate is around 20MB/s.

The Veeam Azure Proxy has been tried with various VM sizes (A, D and F series) and both HDD and SSD storage (using managed disks), but this seems to have no effect on the restore rate. I haven't tried the really large VM sizes however, given the cost. Is there a recommended VM size for the Veeam Azure Proxy?

Does the target storage account type make a difference here? I've tried standard and premium, however premium wouldn't allow a Veeam Azure Proxy to be created on it.

I was particularly surprised to see the restore rate using the Azure-based Veeam B&R server be no faster than the restore rate from the on-prem infrastructure. The Azure to Azure restore test was all done within the same region - US West 2

[oleg.feoktistov]

Hi,

Is there a recommended VM size for the Veeam Azure Proxy?

Depends on amount of data you need to transfer and task slots available on your Azure proxy (1 task slot per machine disk).
Generally, A2 should be enough to process restore data.
If bandwidth is not fully saturated, try increasing the number of concurrent tasks for Azure Proxy corresponding to the VM size (~ 1 task per vCPU core + 2 GB RAM).

Does the target storage account type make a difference here? I've tried standard and premium, however premium wouldn't allow a Veeam Azure Proxy to be created on it.

Shouldn't be the problem on this level. Did you check if VM size of Azure Proxy you're creating is compatible with the storage account you selected?

Best regards,
Oleg