The Windows Server operating system and the Hyper-V Server hypervisor are both NUMA-aware. They will manage thread scheduling and process memory to respect NUMA boundaries wherever possible and to minimize the negative effects of node-spanning when it is inevitable. For virtual machines, this means that guests with vCPU and memory allocations that will fit within a NUMA node will be kept within that NUMA node.

New to Hyper-V Server 2012 is the capability for a virtual machine to be allocated more vCPUs and memory than a typical NUMA node contains. Hyper-V Server handles this by exposing NUMA characteristics to virtual machines. This capability is only one part of the larger subject of virtual machine memory usage, which is the focus of the next section.