Microsoft escalates the ARM race

As Mary-Jo Foley discusses in her blog post, Microsoft's licensing agreement for the ARM architecture can mean big things for Microsoft's future in the mobile computing and smartphone worlds. But being as I'm the datacenter guy my take is more on what it can mean to enterprise computing.

As I have blogged before I really believe in the possibilities presented by large numbers of low-power, high efficiency, CPUs, with good processing power for the future of the datacenter and cloud computing. As Mary-Jo has pointed out on numerous occasions, there have long been rumors of ports of the full-blown Windows operating system to the ARM RISC architecture, though nothing has ever seen the light of day. But let's consider the history of today's Window's operating systems.

The current generation of Windows traces its history back not to MS-DOS, but Windows NT. And those who can remember that far back, will also remember that Windows NT was designed on RISC processors, with the DEC Alpha being a primary NT development platform (after pretty, much everyone gave up on the Intel i860 RISC CPU).  The OS was also ported to the MIPS RISC architecture, and even made it onto the X86 architecture, which eventually overshadowed all other versions, including the Intel Itanium. The point being here is that Microsoft has a long (and recent) history of porting their flagship OS to RISC type architectures, so why would they have a problem porting to ARM?

So while people think about the possibilities of Windows on ARM, be it Windows Phone OS, Windows Embedded, Windows 7, or Windows Server I want to throw one more possibility out there. Microsoft has gotten an architecture license for ARM. While the chances of Microsoft gett8ing into the chip making business probably run between nil and none, the information that the license provides is what's important. Speculation is running rampant about potential uses for dedicated ARM processors by Microsoft, but consider this; Microsoft can put a lot of effort in porting different versions of Windows to the ARM architecture and lots of that effort would overlap, allowing different groups to take what they need and run with it.  But that isn't the direction that Microsoft has indicated they really want to go, fracturing their OS development in too many different directions.

So how about porting Hyper-V to ARM as the first step, and virtualizing absolutely everything that runs on top of it? Hyper-V is pretty thin and efficient, and rather than developing different versions of operating systems for different purposes, a virtualized environment would allow an incredibly flexible utilization of Windows component appropriate to the task.

A virtualization layer sitting on top of a multi-core chip in a smartphone would have plenty of horsepower to deliver any easily conceivable application that a phone user might need (I agree that battery power might be the biggest issue here), and virtualization sitting on top of a box of ARM CPUs could likely meet any datacenter processing need that wasn't focused on raw CPU performance (which is most datacenter needs).

Virtualization and distributed processing are where the datacenter is going; maybe Microsoft really has a good idea on how to lead the way.