Microsoft's Hyper-V virtualization platform today moved out of beta. The RTM version is now available for download, and will be released to Windows Update on July 8. Hyper-V, previously codenamed "Windows Server Virtualization" and "Viridian," was originally due to ship with Longhorn Server—Windows Server 2008—but due to slipping schedules it was delayed to "within 180 days of the launch of Server 2008.". With Server 2008 going RTM at the end of February Microsoft has kept its word.
Hyper-V is being positioned as a part of the Windows Server platform. To ward off any potential antitrust complaints Microsoft offers Windows Server 2008 both with and without Hyper-V, with the latter version a few dollars cheaper, but the intent is clear; Microsoft believes that virtualization is as much as part of the server platform as Active Directory or IIS. Virtualization is a key technology for server consolidation scenarios, as it allows individual server roles to be isolated without causing a proliferation of hardware, reducing maintenance and cutting power and cooling demands in the data center. Microsoft also believes that virtualization will be essential to offering software plus services platforms in the future.
Compared to Microsoft's previous virtualization platform, Hyper-V boasts significant new capabilities—such as the ability to run 64-bit guest OSes and up-to-16-way SMP—along with improved performance. Virtual machine performance is often limited not by processing power but by I/O performance. Hyper-V demonstrates much greater I/O performance than Virtual Server, meaning that it is now practical to run more demanding server tasks on virtual machines. Although more capable than Virtual server, Hyper-V does not live up to its original expectations. Microsoft initially planned to support live migration of virtual machines (so that they could be moved between physical servers without interruption), 64-way SMP, and the ability to dynamically add processor and memory resources to Virtual Machines. These features were cut to allow the product to ship sooner, and are promised for a future version.
Hyper-V's improved performance is a result of its new architecture. Unlike Virtual Server, which relied on the technique of dynamically translating guest OS code whenever that code attempted to do interact directly with the system hardware, Hyper-V uses (and requires) the virtualization hardware found on most AMD processors and high-end Intel parts. Virtual Server ran guest OSes "on top" of the host OS; each guest OS is essentially a thread or process in the host OS. Hyper-V's approach is different; Hyper-V uses a "hypervisor"—a kind of low-level OS that runs directly on the physical hardware, which controls switching between virtual machines. One virtual machine (which must be running Windows Server 2008) called the "root partition" is able to access the hardware directly. The other VMs communicate with the root partition through the hypervisor. To further improve performance, it is possible to write drivers for the guest OSes that allow direct communication with the root partition.
This new design and improved performance puts Hyper-V in competition with VMware's ESX product. ESX is still the more capable and mature of the two—ESX supports more memory and processors, and live migration, for example—but with Hyper-V, Microsoft's virtualization platform is now suitable for many tasks where Virtual Server was too small or too slow. As the delayed features are eventually implemented, this trend will continue, and Hyper-V will be usable for an increasing proportion of virtualization tasks. Hyper-V is considerably cheaper than ESX; Microsoft's approach is to give the core functionality away for free and then charge for VM management tools, whereas ESX is sold in the more conventional way. For those who need the extra features, ESX is still a good option, but Hyper-V is sure to cause some VMware customers to look elsewhere.
No comments on this post yet!
You must be logged in to post a comment.