Physical vs Virtual Proxy

[antuk]

Hi all,

Ive read all the articles but im still not clear what would give better performance in my scenario.

Im running a physical veeam server with the proxy running on itself, and this is backing up a VMware environment over a 1Gb network link.

Would deploying a proxy server on the vmware environment speed up full backups (which take nearly a day to complete), or not since the virtual proxy would still have to send traffic over the same 1Gb network link?

Thanks,
Anthony

[Mike Resseler]

Hi Anthony,

The proxy does a couple of tasks such as
- Retrieving VM data from the production storage
- Compressing
- Deduplicating
- Encrypting
- Sending it to the backup repository (for example, if you run a backup job) or another backup proxy (for example, if you run a replication job)
(this comes from here: https://helpcenter.veeam.com/docs/backu ... tml?ver=95)

Depending on your environment, and how it is all connected to each other, you might get some advantage in using a virtual one. But I still think the network will become a bottleneck. (I assume you see now that the network is the bottleneck in the job statistics?). What you can do (rather easily) is deploy a new VM on your VMware environment and make it a proxy. Then you can test the difference and you can still easily throw away that VM after the tests.

My 2 cents,
Mike

[Andreas Neufert]

Please check out the following
https://bp.veeam.expert/proxy_servers_i ... sport_mode

and the whole Proxy chappter of the best practices guide. It should give you a good overview what to choose and the pro and cons.

[bdufour]

when you say 1gbe - is this over the LAN or WAN, i would assume LAN, but all day to do a full backup seems very long even over a 1gbe network, although we dont know how big this vm is or how saturated the underlying network actually is.

things i would check:
-as mike said - def verify where the bottleneck is according to veeam.
-is there some sort of throttling rule on the network and/or repository in veeam.
-is the job taking long bc the job doesnt have the required resources to start straight away (like too many jobs running at one time).

[antuk]

The full backup is about 2TB of data across just over a dozen VMs, all running in their own jobs. Its 1GB LAN.

The stats are showing
Load: Source 99% > Proxy 47% > Network 0% > Target 0%
Load: Source 99% > Proxy 54% > Network 0% > Target 0%

For the two biggest VMs. Im thinking if i deploy a vm proxy it would move the bottleneck towards proxy/network?

[Mike Resseler]

Antuk,
The bottleneck shows the source. So it seems the issue is with the VMs or hosts. I personally would start looking first what the problem is over there when a backup is running and see what resources are going to the maximum

[antuk]

I read source as changing the proxy would help hence the question.

So ive done what Mike suggested and tested deploying a new VM as the proxy, and im shocked at the difference, its cut 12 hours off the backup time!!

Bottleneck now show as Network as i expected

[neilmurphy65]

Hi Anthony,

I had exeactly the same question as yours some time ago (see vmware-vsphere-f24/newbie-question-hot- ... 10405.html).

Basically, a virtual proxy should speed things up because "the benefit hotadd provides comes from the fact that data is transferred to the proxy server via ESXi storage stack, which is much faster, instead of the network stack.)

I've tried this in my own lab and got a significant performance boost. I've written up my findings on my blog here: https://univirt.wordpress.com/2016/10/1 ... rformance/.

Hope this helps,
neil.