Testing the Performance of ESXi in VMware Player written 7 years ago


What happens to performance when you run VMWare ESXi inside VMWare Player?

Like a circle in a spiral
Like a wheel within a wheel
Never ending or beginning,
On an ever spinning wheel
As the images unwind
Like the circle that you find
In the windmills of your mind

Setup - Hardware

Setup - Tests

  1. To test disk and CPU Performance, video encoding is a pretty good benchmarking tool. To test each platform we would transcode the Big Buck Bunny 1080p Trailer from Quicktime .mov format into the Apple TV2 Preset within Handbrake and note the Average Frame Rate.

  2. To test synthetic (single threaded) CPU Performance, we run SuperPi and take the time taken to calculate PI to 2 Million places.

  3. To test synthetic Disk performance, we run Crystal DiskMark and note down the Read/Write speeds.


Reference Environment 64bit 4 cores 16GB Bare metal.
CPU has 4 cores
One Layer of Virtualisation 32bit 4 cores 4GB Windows 7 running inside VMWare Player on Windows 7. (32bit as this is the only version we had available)
Two Layers of Virtualisation 32bit 4 cores 4GB Windows 7 Running inside ESXi, Running Inside VMWare Player on Windows 7. (32bit as this is the only version we had version available)


System Video(fps) Super Pi(s) Disk IO Read (MB/s) Disc IO Write (MB/s)
Windows 749.0025.116324.7319.329.5622.76229.3218.288.8991.21
Windows 7 in VMWare Player49.0226.875269.9205.316.1014.70234.6224.029.039.01
Windows 7 in VMWare ESXi in VMWare Player50.0729.60962.0752.503.958N/R56.29N/R4.281N/R
Windows 7 in Microsoft VirtualPC†~18††Not Run177.1170.67.8066.863116.894.017.5589.238

† We happened to run the tests by accident in VirtualPC. It’s fair enough to say it’s pants.
†† VirtualPC Bluescreened when encoding the video.
††† Where a test indicates Not Run (N/R) it’s mostly because it was fast approaching 2:30am, and we mostly had to be in work the next day.

VMWare ESXI running inside VMWare Player
VMWare ESXI running inside VMWare Player Performance Graphs

The general gist is that the CPU is virtually unaffected by virtualisation, but disk performance suffers quite a lot. Whether or not it affects day to day performance is hard to say - here the video re-encoding runs in near same time (quicker, even!) inside two levels of virtualisation. Another consideration are the settings used - the disk when run inside ESXi would only mount as an IDE drive, and was not recognised as an SSD: which might have hampered performance. I’m not sure if more tinkering here would have yielded better results.

The tests also indicate it’s well worth researching your virtualisation software: Microsoft VirtualPC had terrible CPU performance; and dire Disk performance in comparison to VMWare.

← previous entry | next entry → Mon 28th May 2012 - 23:11 | no comments | tagged with Computery Stuff

 There are no responses to this entry

Do you want to say something? Use the form below to add your thoughts.

Note if you submit your comments, they'll appear here! yes, right below this. Try and make them relevant eh?

Leave a reply

Name: displayed next to your comment
Email: will not be displayed
URL: to link back to your site
Comments: Markdown syntax allowed

Notes on leaving replies: