What is the typical backup speed for this type of setup?

[TommyB]

Hi,

I'm running VEEAM in the following environment:

Vmware Vsphere v5.1 cluster with two hosts = more than enough horse power, typically the cluster is at 10-15% load
VEEAM runs on a dedicated physical machine, connected via Gbit LAN
Backup proxies are both VMs, one of them is only used as a proxy, the other one is our Blackberry server with very low load.
All Backup jobs, apart from the Blackberry server, have guest file indexing activated

Typical processing rates are around 30MB/s, bottleneck is always the proxy.
The only exception is the Blackberry VM, which has a processing rate of 300 - 500 MB/s

When the backup runs, the backup proxy is at 100% CPU load - adding more cores doesn't change this.

Incremental backups are no problem, as they typically only take some 10 minutes... but if I need to create a full backup of the larger VMs, this would take 8-10 hours.

Is the difference between the Blackberry VM and the other ones only caused by guest file indexing or is there something else to be considered to speed things up?

Regards
Thomas

[dellock6]

As always, what are the bottleneck values of the job involving the large VM? This is the first step to start investigate performances...

[Vitaliy S.]

Hi Thomas,

Since you say that proxy is a bottleneck and CPU usage is maxed out during job run, then you should either add more CPU resources or lower the compression level for the backups jobs.

Thanks!

[veremin]

Additionally, I'm wondering what product version you're currently at and what compression level is specified. Thanks.

[foggy]

VM guest OS indexing processing time typically depends on the number of files stored inside the VM (might take a while, in case huge amount of files. Also, please keep in mind that various processing speed for different VMs is expected.

[TommyB]

Additionally, I'm wondering what product version you're currently at and what compression level is specified. Thanks.

7.0.0.839

Settings are:

Inline dedupe activated and "high" compression - I will reduce this to "optimal" and see how much difference that makes

I also just checked disk read/write speed and since this is quite an old machine (Intel SE7520JR2 Board from back in 2005), it probably will never get above 60MB/s raw write to disk speed.

So probably if I can manage to get rid of any proxy bottleneck, the disk write will be the next one

Thomas

[foggy]

Some of the chain components will always be a bottleneck, so there's always a place for improvement, in every environment.

[veremin]

Inline dedupe activated and "high" compression - I will reduce this to "optimal" and see how much difference that makes

It's, indeed, recommended to use Optimal compression level in order to reduce the load put on proxy CPU resource, as this is the exact purpose the v7 Optimal level was designed for. Thanks.

[TommyB]

Hi,

I just wanted to give some feedback on that topic.

After changing to "optimal", transfer rates for the Microsoft Exchange VM (just as an example) have changed from typically 30mb/s to nearly 100mb/s while the compression ratio has not suffered significantly.

Thomas

[veremin]

Thank you, Thomas, for coming back and updating us with the recent information; appreciated.