Server 2016 Impact on VDI User Experience
At Login VSI, we love new technology and since Microsoft has released Server 2016 yesterday we thought to give you some insights on how upgrading from Server 2012r2 would impact user experience.
If you’d like more info about this release Microsoft is doing live webcast for an intro on Server 2016 later today (October 13, 9 AM Pacific Time (UTC-7)). There are already a lot of people discussing all the new features like Containers, Nano server, storage spaces or the impact of licensing and of course the Microsoft RDS team has added a lot of cool new stuff including:
- OpenGL support in the RemoteFX vGPU adapter.
- GPU pass-through support.
- Broker support for large scale deployments.
- Optimizations for cloud deployments.
- And much more.
We have not been sitting around waiting for the GA. We wanted to know how Server 2016 performed, so we grabbed the RTM build 14393.rs1 two weeks ago and gave it a spin in our labs to see how it would score on user experience. The good news is, other than the HTML5 videos not playing (resolved by running Windows Update) we did not run into any significant issues setting up the test.
We did however discover a definite performance degradation. Preliminary results show that the VSImax for Server 2016 is 15% lower. Let’s see what these numbers mean. First we installed a clean Server 2012R2 and Server 2016 virtual machine and let Windows Update fully patch the systems.
Relative difference between Server 2016 and Server 2012R2
When these tests where completed we did “some” performance tuning on the images using our own Login VSI template in VMware OSOT. This did increase overall performance by a bit as you can see below. The relative difference became smaller because Server 2012R2 barely gains from performance optimizations. So on Server 2016, performance tuning your VDI image is still highly recommended, especially since anti-virus is now included by default.
Relative impact of performance tuning
It is not the first time Microsoft has provided us with an update that has a significant impact on performance. We all remember the tests where we compared Microsoft Office 2010 vs 2013 and discovered a 20% performance hit. Alternatively, who remembers the impact of going from Windows 7 to Windows 10? We saw a 12% drop in performance in systems where CPU is the bottleneck as published here.
Should this stop you from upgrading? No, absolutely not, this is simply a reminder that the performance impact should never be underestimated whether the change is small or large, especially when it’s so easy to validate.
Please note: These tests have been executed in a lab environment where we know CPU is the bottleneck. Results may vary when testing Server 2016 in your own environment and when adding your own applications and policies into the test.
With #VDILikeAPro we believe sharing is caring, therefore all of these results are available for the world to view, use and analyze. You can download a zipfile here that contains: Login VSI logfiles, VMware ESXtop performance logs and some additional performance charts in a XLS file. Feel free to use this information to your liking but please keep in mind the policy.
Note: The White Paper is available for download.