Customers often ask “Given a hardware load-out X, how many Ys can I get/run/host?”

The frustrating answer always starts with “…it depends…”. We caveat this way not because we want to frustrate, but because it is true. Many teams will go on to say “We have tested the following in our labs and gotten the displayed results…”

1 X = thingy

2 X = more thingies

3X = many more thingies

While accurate, not super-helpful.

The truth is that to do good perf planning for Hyper-V you have to run some tests.

Run them using your actual production load (converted to Virtual Machines) in a test environment.

TIP: you can download the free VHD version of SCVMM, then run it as a VM to convert your production machines to “test” virtual machines.

Then play with your assumptions and tweak things higher and lower and to your design tolerance and actually observe how perf goes.

Add an overhead/forgot-to-test percentage, done. 

So, how to do that for Hyper-V? Here are some perf testing resources to help you: