WAN Accelerator question

[mfc]

We have replication jobs that replicate VMs from our HQ site to our DR site over a 10Mbps WAN. The Veeam B&R server (at DR site) and proxy/WANaccel (at HQ site) are both physical machines with dual 6-core CPUs/16GB RAM, and the WAN caches are on SSD drives.

We are using significantly less bandwidth (about 25%, we used to saturate), but our repl. times have not really improved because the source WAN accelerator is limited to one job at a time. We'd like to shorten the repl. window, I have a couple ideas I want to run by you all:

1. Add more physical proxies/WANaccels at the source site. For cost reasons, this would have to be more workstation-class hardware (with SSDs).
2. Add virtual proxies/WANaccels at the source site. This is essentially "free", but the downside is our VMware datastore resides on 15K SAS (not SSD). Also I'm slightly concerned about adding IOPS to our SAN, since replica jobs run during business hours.

[Vitaliy S.]

Hi Mike,

What does bottleneck statistics show for your replication jobs and how many VMs you're replicating to the offsite location?

Thanks!

[mfc]

The bottleneck is normally Target WAN, with Source being second. For example, most jobs usually show something like 85-Source, 94-Target WAN on the bottleneck stats.

We're currently replicating 21 VMs, with 3 of them being fairly large Exchange and SQL servers.

[Gostev]

Are you using v7 or v8? Most v7 installs, and especially v8 should be fully saturating 10 Mbps.
25% of 10 Mbps is waaay too slow processing rate, you need to troubleshoot this with support.