Windows Takes On the High Performance Computing Market

It couldn't have come at a more opportune time: With Wall Street reeling under the weight of a proposed $700 billion federal bailout, Microsoft on Monday announced the completion of Windows High Performance Computing (HPC) 2008. Not surprisingly, given the current climate, Microsoft is touting Windows HPC 2008's risk analysis abilities.

I wonder if this message was carefully crafted as this year's financial market problems became more obvious. But regardless of the advantageous message at the HPC launch event this week, I think it's pretty clear that this technology has a solid future in environments far beyond Wall Street.

Windows HPC 2008 is classic Microsoft: Take a technology that was previously expensive, exclusive, and niche and bring it to the masses at ever-more affordable prices. What it's really about is compute clusters of all sizes (the previous version was actually called Windows Compute Cluster Server), a highly scalable job scheduler, and parallel processing. And when you look at the product in that light, you can see that it's only a matter of time before these capabilities begin trickling down the Windows ecosystem.

Windows HPC 2008 is a 64-bit solution, of course, but it runs on the mainstream x64 hardware platform, and not some esoteric chipset you've never heard of. This has two benefits. First, it means that existing applications and hardware devices will just work (though it should be noted that HPC 2008 is not designed to be used as a general-purpose server OS). Second, moving HPC's high-end capabilities downstream will be all the easier.

Windows HPC 2008 is familiar. Its management model is identical to that of other Windows Server products, and based on System Center technologies, so it's easy to get up and running. This too, is classic Microsoft, a management democratization of sorts that lets any Windows administrator or IT professional easily seek previously out-of-reach employment opportunities.

Windows HPC 2008 is inexpensive. It's possible to create a small HPC cluster for $5000 to $10,000, a price that Microsoft notes is about 10,000 times less expensive than the cost of similar computing power in 1991. (Microsoft charges $475 per node.) A basic compute cluster will include a head node server and then some number of computer nodes. The head node connects the cluster to your broader internal network, but communication within the cluster occurs through a separate, private network.

Windows HPC 2008 is for developers. Thanks to integration with Visual Studio 2008, it's possible for developers to write HPC solutions in record time and take advantage of the product's unique functionality. Developers can also easily port their HPC solutions from more expensive and complex solutions to Windows HPC. And Windows HPC 2008 is fully compatible with the standard Message Passing Interface (MPI) message-passing API and specification, which is widely used by existing HPC clusters.

It's pretty clear that Microsoft is serious about taking high-performance computing mainstream with Windows HPC Server 2008. If you've previously found this type of solution to be too complex or too expensive, now is the time to take another look.