Hyper-V reproducible results for performance testing

I am running some performance tests on a Hyper-V machines. Basically, I want to see how the performance of the product varies with new product releases. The nature of the product is, among others, file-intensive, let's say an file-compression product.

Is there a (known, documented) way to configure Hyper-V/Windows Server to minimize impact on my test runs? For example, no cache is done outside and/or inside a Hyper-V virtual machine.

I restore the previous snapshot before every test run.