Adventures in Benchmarking

So far in my experimenting with Hyper-V, I've been impressed by the speed of the platform (once the autodetection of the HAL was turned on and I was able to install the Integration Services successfully). The question arises... how fast, exactly, *is* Hyper-V?

One way to answer that question is to run some benchmark software. It turns out that there's not a whole lot of benchmarking software available for Server 2008, so I found myself looking for tools that would work with Vista and, in some cases, hacking them so they'd work on Server 2008. The two suites I found that could be made to work included products from SiSoft and FutureMark. The FutureMark product required a patch that I found on the Web; the SiSoft tool worked without patching.

Now I was so curious that I went ahead and ran some tests even though they're not scientifically accurate, because the child OS that I've been using is running 32-bit Server 2008 and the parent OS on my Hyper-V machine is running 64-bit Server 2008. What I really need to do is to create a 64-bit child VM and then compare apples to apples. That's on the schedule for next week... but meanwhile, I can tell you that even in my apples-to-oranges informal tests, the child OS clocks in at between 75 and 85 percent of the parent OS on a mix of benchmarks. That's pretty darn good in my book - I shudder to think what the same measurements might look like with Virtual Server 2005! More valid results will be forthcoming in future posts.

Recent posts:

Hyper-V Manager and Helpless Help

Hyper-V Manager on Vista – final thoughts

Server 2008 Hyper-V: Managing Hyper-V from Vista, Part I