Different transfer rate speed on replicated VMs

[icebun]

I was doing a couple of quick tests which were as follows:-

20GB VM - initial replication was done over a 6Mb WAN. Using CBT thereafter I was averaging 15MBs

65GB VM - Pre staged onto removable media, locally copied to DR SAN and resynced. Using CBT thereafter, I was averaging 176MB

Both these VMs have minimal data changes (they are test ones).

Is there any reason why I should see different speeds?

By the way I am using Virtual Appliance mode.

One more thing, so that I understand the mechanics, using this mode means that the source ESX host communicates with destination ESX host and traffica is via the Nics linked to the service console. If that is correct, are there any tips with regard to allocating perhaps more memory to the SC or dedicated vSwitch for the replicated trafffic?

Thanks

[Vitaliy S.]

Hello,

Can you tell us whether those VMs where you got 15MBs and 176MBs on incremental runs reside on the same datastore/the same host? What type of VM were you replicating, were you using the same destination for both incremental test runs? If you're using Virtual Appliance mode try giving 4 vCPU to your VM and change compression value to highest level and re-run both jobs. It should give you better performance.

Thank you!

[icebun]

Hi Vitaliy,

Thanks for the super quick response!

Ok both VMs are W2k3 servers.

Each VM resides on a different datastore source, but both VMs are from same source Host. The destination datastore and host are the same.

I will make those changes as recommended but can you please clarify the original question relating to the traffic from the hosts and service console Nics? I am trying to undersand the performance hit if any on the host servers as well.

Can I also ask the relevance of the datastore location question?

Thanks.

[Vitaliy S.]

Ok, the speed difference may relate to many variables in your infrastructure, for example the particular datastore may be the main bottleneck for your speed issues. I would try to backup other VMs from both datastores and see whether there is any difference comparing to your initial results.

In case of Virtual Appliance mode usage Veeam uses SCSI hot-add capability of ESX to attach disks of a backed up VM to the Veeam Backup & Replication VM. In this mode, VM data is retrieved directly from storage through the ESX I/O stack, instead of going through the network stack, which improves performance. After that in case of network attached storage (for example) all the traffic flows to your destination storage via NIC.

Thank you!