This might seem like a no-brainer – optimizing your VDI images should make them run leaner and perform better, all else being equal. That being said, there’s not always a ton of data backing that up.
I recently had a request to test a specific configuration with unoptimized images with Windows 10 and Office 2016. I thought, while I was at it, I wanted to get some comparative performance metrics versus an optimized image.
- Login VSI 4.1.4
- VMware Horizon 7.0.2
- 2x SimpliVity OmniStack nodes running vSphere 6.0
- 300x Windows 10 full clone VMs with Office 2016 + Login VSI apps installed
- Knowledge Worker Login VSI workload
- VMware OS Optimization Tool
Now, to preface this, I want to point out what I’m actually looking for in the results. VSIbase should be consistent for all tests, as no workload, aside from the idle VMs, is running. I want to look for upward trends in response time and VSImax average as my key independent variable. So here are our baseline unoptimized runs.
So the average VSImax average for these two runs is 1044ms. That’ll be our baseline for comparison.
For this test, I deleted my unoptimized VM pool, optimized the image using the VMware OS Optimization Tool and the included Windows 10 Horizon Air Hybrid profile.
As anticipated, the VSIbase value is pretty consistent across all four runs. However, in contrast to that, the average VSImax average for these two runs is 901ms, which is an improvement of 13.7% over the unoptimized image. When you talk about the scale at which most VDI is deployed, that kind of difference is huge.
So remember, kids – always optimize your images prior to deployment.