With Windows Server 2016 we have gained some very welcome capabilities to do cost effective VDI deployments using all in box technologies. The main areas of improvement are in storage, RemoteFX and with Discrete Device Assignment for hardware pass-through to the VM. Let’s take a look at what’s possible now and think out loud on what solutions are possible as well as their benefits and drawbacks.

With Windows Server 2012 Hyper-V became truly NUMA aware.  A virtual NUMA topology is presented to the guest operating system. By default, the virtual NUMA topology is optimized by matching the NUMA topology of physical host. This enables Hyper-V to get the optimal performance for virtual machines with high performance, NUMA aware workloads where large numbers of vCPUs and lots of memory come into play. A great and well known example of this is SQL Server.

The need or perceived need for hard CPU processor affinity stems from a desire to offer the best possible guaranteed performance.  The use cases for this do exist but the problems they try to solve or the needs they try to meet might be better served by a different design or architecture such as dedicated hardware. This is especially true when this requirement is limited to a single or only a few virtual machines needing lots of resources and high performance that are mixed into an environment where maximum density is a requirement. In such cases, the loss of flexibility by the Hyper-V CPU scheduler in regards to selecting where to source the time slices of CPU cycles is detrimental. The high performance requirements of such VMs also means turning of NUMA spanning. Combining processor affinity and high performance with maximum virtual machine density is a complex order to fulfill, no matter what.