As you have likely heard by now, SP1 for Windows Server 2008 R2 Hyper-V will soon be released (in the first quarter of 2011). Yesterday at TechEd Europe, Microsoft announced the results of recent joint scalability testing by Microsoft and Citrix on the scalability of Hyper-V SP1 and Dynamic memory with Hyper-V in Windows Server 2008 R2 SP1.
Brad Anderson (Corporate Vice President, Management and Security) announced the results in his keynote from Berlin"...with Service Pack 1 you will see a 40% improvement in the density that we enable on Hyper-V in a VDI environment".
Michale Kleef, Senior Technical Product Manager went through a demo with Hyper-V SP1, Dynamic Memory, Remote FX and Citrix XenDesktop. You can watch the 10 minute video on at the TechEd Europe site here.
This section of the keynote starts around 1 hour and 10 minutes into the video and goes on for 10-11 minutes.
Michael also blogged about the results. Here is a sampling of Michael's post -
As Brad Anderson and I discussed at the TechEd Europe keynote today, Dynamic Memory, a new feature in Windows Server 2008 R2 SP1, can increase Virtual Desktop Infrastructure (VDI) densities by 40% compared to Hyper-V in Windows Server 2008 R2 and also well above a leading industry solution. It's also not just a benefit to VDI.
...
Our primary goal was simply to understand how Dynamic Memory influences the memory ceiling to density, and realistically, by how much. Secondary goals were to understand how XenDesktop 4 functioned with Hyper-V R2SP1 and its different approach to storage using Citrix Provisioning Services.
This testing demonstrated that, with SP1 enabled, Hyper-V density with XenDesktop increased from 85 Windows 7 VMs to 120 Windows 7 VMs.
By allowing Dynamic Memory to take control of memory allocation, this total load on the single server test up to 120VMs with a 40% increase in density. Each VM averaged around 700MB RAM running the LoginVSI workload and these results were consistently confirmed on the Dell blade testing also. When we scaled this out on the complete Dell RA, we took a reference architecture that previously ran on 12 blades, down to 8, with an easily calculable corresponding drop in cost per user/VM.