The release of Windows Server 2008 R2 Hyper-V is eagerly anticipated, due in large part to the many excellent new features coming with this release. The following information provides an overview of the feature set (as announced) that we can expect to have at RTM.
This is probably the most compelling new feature and certainly the most anticipated in Windows Server 2008 R2 Hyper-V. In a nutshell, live migration provides the ability to move running VMs from one physical host to another (currently, within a single cluster) without any service interruption or any downtime as perceived by end-users. I’ll detail more about live migration vs. Quick Migration and the requirements for each in another post.
We can now hot-add / remove storage! This feature allows the addition and removal of both Virtual Hard Drive (VHD) files and pass-through disks to the existing SCSI controllers of VMs - while a VM is running! Much needed and much appreciated, Microsoft! Thanks…
Note: Hot add and removal of storage requires the Hyper-V Integration Services supplied with Windows Server 2008 R2 to be installed in the guest operating system. I think this is a small price to pay and something that you should really be doing anyway for the best functionality with your VMs.
Have you ever wanted to run a 32-processor SQL Server 2008 Hyper-V VM but were frustrated with the limitation of 24 logical processors provided through Windows Server 2008 Hyper-V? Well, this is going to make your day then…
2008 R2 Hyper-V will now support up to 32 logical processor cores. It also introduces support for Second-Level Address Translation (SLAT) and CPU Core Parking.
Funny Note: SLAT is not the same as the South Lake Union Trolley with its unfortunate acronym. J
Disclaimer: This may only be humorous to readers, local to the Seattle area.
Quote from Microsoft: “SLAT uses special CPU functionality available in Intel processors that support Extended Page tables and AMD processors that support Rapid Virtualization Indexing to carry out some VM memory management functions that reduce the overhead of translating guest physical address to real physical addresses. This significantly reduces Hypervisor CPU time and saves memory for each VM, allowing the physical computer to do more work while utilizing fewer system resources. CPU Core Parking enables power savings by scheduling VM execution on only some of a server’s CPU cores and placing the rest in a sleep state.”
Finally, a couple of networking capabilities from earlier Windows releases are making their way into the Virtualization stack (along with one feature new for just the virtual world)…
The first feature extension is Jumbo Frames support in your VMs. This feature enables virtual machines to use Jumbo Frames up to 9014 bytes in size if the underlying physical network supports it. Supporting Jumbo frames reduces the network stack overhead incurred per byte and increases throughput. In addition, there is also a significant reduction of CPU utilization due to the fewer number of calls from the network stack to the network driver.
TCP Chimney, which allows the offloading of TCP/IP processing to the network hardware, has also been extended to work in the virtual world. TCP Chimney improves VM performance by allowing the VM to offload network processing to hardware, especially with networks over 1 Gigabit. This feature will prove of particular value when VMs are involved in large data transfers.
As for the new feature, the Virtual Machine Queue (VMQ) feature allows the NICs of the physical computer to use DMA to place the contents of packets directly into VM memory, increasing I/O performance.
Cluster Shared Volumes (CSV) is another of the big additions to Server 2008 R2. While it is commonly misunderstood to be a requirement for Hyper-V live migration (it is not required; other third-party products can also provide the shared volume support needed to permit live migration to function), it is perhaps the best solution as it is in the box with R2 and it simply works (and it’s free).
CSV is a significant departure from the last 13+ years of cluster disk architecture from Microsoft, in that it now provides access to a single volume from multiple cluster nodes simultaneously. I’ll get into the details of CSV in another blog post, but suffice it for now to say that CSV greatly simplifies the process of working with SAN volumes in large clusters (no more drive letter vs. volume GUID deliberations) and enables us to use live migration without additional third-party software.