High Performance

Microsoft has claimed the lead in server software market share for a good while now, and Windows Server is pretty ubiquitous. But one area in which most of us probably don't think of Windows as the logical platform is high-performance server software for heavy, complex calculations. That may well be about to change. Microsoft has sent its Windows Compute Cluster Server 2003 to manufacturing, and signs are that this release is poised to jump Microsoft into the regular conversation when we talk about HPC environments. The move toward clustering and, to an extent, away from supercomputers plays totally to Microsoft's advantage; they have the code to run servers on smaller boxes, and linking that up for clustering is a task well-suited for the company. The software should give them easier entry into markets such as science-based computing, high-end mathematics, biochemistry, engineering, and other technically oriented HPC environments. The University of Tennessee is already cranking a system using Compute Cluster Server, and the speed with which it's running is expected to get the UT cluster onto the Top 500 Supercomputers list.

Does this have any meaning to you, the small- and midmarket server manager? Well, sure. Microsoft can eventually reflect the speed boosts in its Compute Cluster Server code back into the flagship server software -- most likely not until we see a couple of iterations of Longhorn Server. But it'll get there. It's irresistible for Microsoft to avoid boosting its core market if it can do so, and no doubt it has found, or will find, some areas of Compute Cluster that it can migrate to Windows Server proper to help speed and efficiency. Microsoft looks at its server software as one large interconnected entity, and the cluster software is no different. You may not be sequencing the genome of the iguana or anything like that, but you will eventually benefit from the software that does -- count on it.