Optimizing planned failover performance

[StefanNL]

Hi all,

We are planning on using Veeam for a 600VM datacenter migration.
We have a 2x10Gb connection between the DC's and network tests between the Veeam proxies show close to linespeed on the netwerk, the distance is very short.

I have learned some things already that have reduced the incremental replication time from a small sample group.
This group contains 10VM's and has a total size of 905GB
Full replication with two jobs takes 15 minutes, incrementals take 11 minutes using two big proxies on the source and two big proxies on the target side.
The full replication reaches up to 1GB/s during the initial full replication, so we are very pleased with this.

However, then things get a bit complicated.
The planned failover step takes 52 minutes for the same 10VM's.
Now we need to migrate up to 100-150VM's per weekend but that is going to take more than 10 hours with this method, this is unworkable.

We divided the VM's into 5 jobs of two VM's and configured them all on a single proxy.
This brings the planned failover duration down to 40 minutes (from the original 52) but this is still way to long.

So here comes my first question:
How do I configure the environment to perform in an optimal manner for reducing the duration of the planned failover feature of Veeam?

I would accept the planned failover to take twice as long as a normal incremental replication (because it does two of these) but not almost four times as long as it does now.

And secondly:
Why doesn't the planned failover use multiple parallel tasks per proxy like a normal replication job allows for?

Regards,
Stefan

[StefanNL]

Just to clarify.
We have tested a few options.

one proxy per site with 1 job in total that contains 10 VM's does a full in 19 minutes, incremental in 18 minutes, no planned failover was tested in this configuration.
two proxies per site with 2 jobs in total that contain 5 VM's per job does a full in 15 minutes, incremental in 11 minutes and a planned failover in 52 minutes.
A single proxy per site with 5 jobs in total that contain 2 VM's per job does a full in 18 minutes, incremental in 18 minutes and a planned failover in 40 minutes.

Extrapolating these numbers to 100-150VM's in a planned failover is no fun and any guidance is much appreciated.
The goal is to get a close to a 20-25 minute planned failover time for the test workload as this would be 2x the incremental replication in the optimal setup we have tested.