Hyper-V in this release continues its maturation as a compelling hypervisor platform. Reports indicate that in pre-preview testing, Hyper-V in Windows Server 8 supports up to 160 logical processors, 2 TB of RAM, 32 virtual processors and 512 GB of memory for virtual machines, along with support for guest NUMA and an end of the virtual-to-logical processor ratios.

This all comes into play when you consider scaling up -- especially in a cloud scenario. Why does NUMA, or non-uniform memory access, matter? Essentially, as a developer, you want to make sure processors are scheduling threads locally and allocating memory as best they can. You want to avoid crossing node boundaries to avoid latency, slow caching and other performance-impacting symptoms, since allocation and latency depends on the memory location relative to a processor.

High-performance applications detect NUMA and communicate with the OS. With guest NUMA, Hyper-V presents NUMA topology within a virtual machine, allowing the guest OS and applications to make intelligent NUMA decisions about thread and memory allocations.

Indeed, virtualization in Windows Server 8 even one-ups physical hosting. With new predictive failure-analysis capabilities, through firmware or processor signaling, the operating system can find out when errors are occurring. In this way, virtualized workloads actually exceed physical capabilities since the OS can do some intelligent page analysis when memory is virtualized and not physically allocated. In essence, the very fact that you're running Hyper-V in the first place means you have an extra layer of resiliency that isn't present with physical hardware.