Veeam restore speed

[SE-1]

Hello,

We noticed that the restore speeds with veeam in combination with DD are extremely slow.

We achieve speeds of 5MB/s on a 10Gbit network on a DD2500 with 1 expansion (27 disks).
In the system stats of the DD we see on network out 2MB/s to 15MB/s

As a test we have installed vRanger from DELL, we take a backup of the same VM to the same DD and launch a restore at the same time.
The vRanger restore speed is around 80-90MB/s and we see in the system stats of the DD network out between 120-130MB/S, which is result that you would expect.

Support Case ID: 00962140

Why is veeam restoring so slow?

Thank you

[Gostev]

Hello, are you comparing the restore speed from a full backup of a newly created job containing a single VM with a single disk, and with parallel processing disabled in the global settings? This would be the closest set up to vRanger. Thing is, Veeam has a lot of advanced functionality aimed at raw storage that deduplicating storage may not "like", because they expect legacy streaming workload (just as if you were to back up to tape). Thanks!

[SE-1]

Hello,

Enable parallell processing is enabled under IO/Control panel and enable storage latency is disabled.

The VM restored has 4 virtual disks

For example: I have started the restore of a vm.

It started @ 6.47AM. The vCenter message telling that the vm is registred is at 7.00AM, which is 13 minutes.

The step: preparing disks for restore happend at 7:09, which is 22 minutes after the restore job started.

It starts moving data at 7:15, which is 30 minutes after the job initally started.

In the Job log the process rate = 1MB/s, in the statistics tab the restore rate is 20MB/s.

Why does it takes 30 Minutes to even start moving data?

[chris.childerhose]

I have had similar issues as we use Data Domain for backup target and are moving from a DD670 to DD2500.

I did a comparison with VDPA which is the VMware Data Protection Advanced backup and a restore that took 16 hours (SQL Server) in Veeam took 4 hours in VDPA. I believe it is the number of streams for the restore that is the culprit where Veeam only does 1 and the other products can do multiple.

Think there is some tweaking still needed or allowing multiple streams for the restore process. Otherwise we have moved to using storage snapshots on Nimble as our first line of restore before restoring an entire VM which is painful.