Well, it’s finally time for part three: test results. When I began my look at FVP, I had decided I wasn’t going to test the actual performance gains because my lab is not a good representation of a real environment. It’s good for “logical” testing; for instance, in this case I wanted to walk through the install and see the GUI. But since it’s nested and run on consumer hardware, I didn’t figure it would be fair to test a product in a situation that didn’t accurately represent the intended market.
Curiosity got the best of me, however, and I ran some tests just for fun. I ended up getting really great results and I really want to share them! I just want it to be clear that your mileage may vary drastically in either direction because of my setup. OK, disclaimer over; on to the good stuff!
VMware IO Analyzer
To get some test results, I went ahead and ran some benchmarks using VMware’s IO Analyzer tool. FYI, It runs IOMeter on the back end, in case you haven’t played with it yourself. I ran the tests a few times to get some averages, and what I’m showing below are not actual averages, but are results that I feel are “about” average across the multiple tests.
Here, we see that I ran a 64k workload, simulating perhaps a database server. I ran the test once without FVP, enabled FVP, and ran the test again. What we see is around 1760 IOPS before, and 2825 IOPS once FVP is enabled. A marked improvement!
I was also able to grab some charts of these runs, which I’ve provided below, showing the improvement. Long story short, toggling FVP on for this VM resulted in a 50 percent decrease in IO latency, and a 60 percent increase in IOPS!
All things considered, I really enjoyed using FVP. The installation was simple enough, as we’ve come to expect these days. Administration is done from a plugin that is a part of the Web Client so that you can manage from (DON’T SAY IT!!!) a single pane of glass! 🙂 The results are measurable and immediate. Overall, I would definitely recommend FVP for someone looking to squeeze additional performance out of their existing array.
This post is not sponsored by PernixData and I have not received any compensation or benefit from them for writing it. The information contained is purely from my own testing and my own opinions. I obtained an FVP demo license the same way a potential customer would, and set this up as if it were a POC for a real customer.