Hardware req advice for v6 backup proxy server (vCPU count)

[Pretzeldude]

Hello,

I've setup a new Veeam server as a VM (VMWare Esxi) with 8GB RAM and 2 vCPUs. Windows 2008 R2 Standard.

I've noticed that when I run backup jobs (even though they are successful) that the veeam agent is maxing out the CPU until the job finishes. The machine hardly responds to anything else until the job is done.

Any advice? Is this a bug or have I not applied enough resources for a backup server?

Thanks,
Ted

[foggy]

Ted, do you have your Veeam B&R server used as a backup proxy for your jobs? Actually, CPU saturation is quite normal in the case where the source storage performance allows very fast data retrieval (please look at this topic for some explanations). You could add more vCPUs to this VM to handle the load.

[Pretzeldude]

That was helpful. Thanks. I am using the B&R server as the proxy at this point. My jobs seem to be getting about 87MB/s throughput and the log says that the bottleneck is "source".

[Gostev]

So, how much CPU resources would you expect deduplication and compression of 87MB of data each second to take? This is pretty massive data stream, and compression is very CPU heavy task. Try simply zipping some big file (such as ISO) with Windows Explorer on the same server, and watch the CPU load.

As per our system requirements, 2 vCPU is a minimum. I always recommend going with at least 4 vCPUs with fast backup infrastructure.

Thanks!

[stevenrodenburg1]

My jobs seem to be getting about 87MB/s throughput and the log says that the bottleneck is "source".

I get that in our v6 Lab environment as well. Always "source" as the black sheep
The Proxies now has 4 dedicated vCPU's (Xeon E3-1270 QC, one of the fastest per-core performing Xeons at the moment) but the bottleneck is still "source". Speeds hover between 85 and 90MB/s for a full backup.
Veeam just has an insatiable hunger. It always want more more more!

All four CPU's hit the ceiling during jobs. In Virtual appliance mode, one must watch out for situations where the Veeam virtual appliance VM's monopolize a hypervisor with their greed

[Gostev]

Upgrade to vSphere 5 and go 8 vCPU for proxy VM, this will satisfy Veeam's hunger

By the way, for those running jobs during production hours - remember that you can always reduce resource allocation for backup proxy VM to make sure other VMs on the same host are not affected by the "noisy neighbor".

[bsousapt]

Hello,

I've setup a new Veeam server as a VM (VMWare Esxi) with 8GB RAM and 2 vCPUs. Windows 2008 R2 Standard.

I've noticed that when I run backup jobs (even though they are successful) that the veeam agent is maxing out the CPU until the job finishes. The machine hardly responds to anything else until the job is done.

Any advice? Is this a bug or have I not applied enough resources for a backup server?

Thanks,
Ted

You could try to play around with compression levels and/or disable deduplication.
Like always in order to get some you will lose some, akA trade-offs

Bruno

[gvinpin]

All depends if you are doing de-duplication and compression or not.
My v6 proxy server using 8GB SAN to access VM's and 10GB LAN to deposit data to Data Domain appliance.
CPU is about 1-2% utilisation on it, with 4 jobs running at the same time.
I am not using any Veeam compression or de-duplication, leaving the job to Data Domain appliance.

[Moebius]

[merged]

I just deployed a VBR proxy on a remote branch office datacenter. The datacenter consists in a single ESXi host with local storage. The remote backup repository is a Synology DS412+ NAS configured as a single iSCSI LUN. The VBR proxy is a Windows 2008 R2 vm configured with 2 vCPU and 4GB vRAM.

The backup jobs ran flawlessly with a decent throughput (50 MB/s for the full backups). However, I was suprised seeing the CPUs stuck to 100% for the whole job run. These were full jobs, so there was no CPU load due to the synthetic building phase.
I have never seen a similar behavior on the main Veeam server (console and proxy). The remote proxy has nothing installed on: it's just a plain vanilla Windows with only the proxy service deployed by the Veeam server.
Is such a high CPU utilization to be expected for a VBR proxy?

[Moebius]

Thanks for moving my post to this existing thread, I hadn't noticed it.
I made a typo in my post. Actually the proxy has 4 vCPU and not 2. However, the veeam agent is maxing them out.
It makes sense that deduplication and compression take a heavy toll on the cpu's when the data flow from the source is good. Out of curiosity, I will run a test job with no compression and no dedupe.

[Vitaliy S.]

Yes, the more CPU resources you have on your proxy server, the more concurrent jobs you will be able to run via this proxy server, reducing the compression level will also lower the CPU load on this proxy.