Dedup the backup server as a VM

[kjo@deif.com]

Has anyone here tried installing Veeam in a VM and then deduping the VM VHD files?

Pros:
- Fast clone would work since the VHDs are deduped and not the volume inside the VM.
- If you create multiple VHDs under 4TB in size and combine them inside with storage spaces, then you would be able to dedup servers that are larger than 4TB.
- You can span the VM VHDs over multiple 64TB deduped volumes.

Cons:
- VHD overhead. (But if you have large servers you want to dedup, this might be worth it)
- Virtualization would decrease performance.
- How staple is it?

[PetrM]

Hi Kim!

I think data read speed from backup file might be dramatically affected in this case (anyway, it's always better to test and compare results in a specific environment)
and overall duration of restore operation would be increased.

What about to take a look at data compression and deduplication job settings? (may be to use values higher than default)
By the way, advanced ReFS integration allows multiple full backups on disk to share the same data blocks and you will have spaceless fulls.

And fast clone should work regardless of VHD dedupe or am I missing something from use case description?

Thanks!

[nmdange]

I run Veeam in a VM in the manner you describe and don't have any performance issues compared to before. I haven't benchmarked the config but both before and after the bottleneck is the source system so it's "fast enough". However I don't currently use dedupe since I have more than enough space without it.

This article has some good info on how to do it, while it's for DPM the principles are the same
https://charbelnemnom.com/2016/10/how-t ... p-storage/

[kjo@deif.com]

And fast clone should work regardless of VHD dedupe or am I missing something from use case description?

Hi Petr

If you enable dedup on the server where Veeam is installed, then fast clone will no longer work. There is an experimental feature where you can enable it, but the performance of that is very poor.

Right now we're setting up a second server with this setup and then we will see how it goes.

[PetrM]

Hi Kim!

I think you're talking about this limitation:
"Fast Clone and Windows data deduplication cannot be used simultaneously. Thus, if you target a backup job to a repository supporting Fast Clone and enable Windows data deduplication, the Fast Clone technology will not be used for this job."

I'd expect that it should work as long as VHDs are deduped but not the volumes at the level of guest.
I asked it just to be sure that we're on the same page (it seems you have the same understanding according to the first message).

Ok, let's wait for results of your test.

Thanks!

[kjo@deif.com]

Just an update. I have tried it and I would not recommend it for these reasons:
- It adds a lot of complexity.
- The deduped VHDs can expand, this can cause issues and you have to juggle space two places.
- Dedup ratio is not the best, this might be because of misaligned VHD blocks vs Veeam blocks or meta data.

I think I will try a local object storage with dedup instead, from I have read and tested, it should give me the performance and storage savings I want.