WAN acceleration. What am I doing wrong ?

[jgl75]

Hello!

I'm busy testing the Wan Accelerator but I'm getting mixed results.
I've followed the procedure explained here : http://forums.veeam.com/viewtopic.php?f=2&t=17423
to seed my backups.

I have 3 VMs of different size (55Go, 150Go and 1.3To).
The wan accelerator works brilliantly with the two smaller ones but I can't make it work at an acceptable speed for the large one.
With the wan accelerator, the jobs takes more than 24 hours and the next occurrence of the job interrupt the transfer.

If I'm not using the wan accelerator, the job finishes successfully in 25 minutes.

What can I do to fix this...

I have an uplink of 6 mbits. Tests are done on a vSphere 5.1 host.

Thanks in advance for your help !

Jean

[veremin]

Hi, Jean.

Can you shed a little more light on your deployment? How did you specify Wan Accelerators? What is the size of the cache at both ends? What are the bottleneck statistics for this job? What speed do you see at both WAN and direct methods? Can you confirm that seeding was performed successfully, etc.?

Thanks.

[jgl75]

- The WAN accelerator is defined in the Data Transfer tab of the copy job (local Veeam server in the source - destination server in the destination).
- The cache size is 300Go at the source and 350Go at the destination.
- Here are the bottleneck stats using the wan accelerator: 29/09/2013 06:02:16 :: Load: Source 10% > Source WAN 34% > Network 39% > Target WAN 35% > Target 2%
- Processing rate is 850Kb/s (again with the wan accelerator)
- Stats without the wan accelerator : 01/10/2013 20:07:01 :: Load: Source 46% > Proxy 20% > Network 12% > Target 49%
- Processing rate without the wan accelerator : 11MB/s

Seeding seems to be alright. I managed to map the backup without issue.

Thanks !

[foggy]

As a test, you could perform the same backup copy job locally to populate WAN cache first, then copy cache to the remote site, and see whether this helps.

[jgl75]

I was expecting that answer.
I'll try it as soon as I can.

Thanks !

[foggy]

Keep us updated. Probably it is the global cache population that takes so long for the large VM during the first job run.

[jgl75]

Is it possible to prevent any backup operation as long as the copy job is not finished ?

[foggy]

You can simply disable backup jobs until the backup copy is over.