There has been a constant background hype around Microsoft’s Compute Cluster Server 2003 since SC’06, and yesterday’s announcement from SGI got me to thinking about the Microsoft strategy again.
I am all for trying whatever it takes to get more people using HPC. And anything new takes a while to adopt. All good.
I am a little suspicious, however, of the hype surrounding the purported rapid adoption of Microsoft’s Compute Cluster Server. The Register reported during SC’06 that Microsoft itself is claiming great success on the high end of HPC:
More than half of the clusters sold running Windows Compute Cluster Server (CCS) 2003 have more than 1,000 processors, Microsoft marketing director Shawn Hansen told us today at the Supercomputing event.
But I’m having a hard time nailing that down, and a quick Google today didn’t turn up other instances of the company repeating that claim (their CCS home page does, however, link prominently to the article).
The company’s flagship installation, the Dell PowerEdge-based system at NCSA which placed 130th on the June 2006 Top500 list, was that list’s sole CCS system.
By the November 2006 list that system had been upgraded from 900 PowerEdge 1855′s to to 2048 PowerEdge 1955′s and switched from Microsoft to Redhat. There are zero entries for CCS on the current Top500 list (interestingly there are 3 for Mac OS X).
I know there is lots of HPC done in the world that isn’t represented on the list; some of it may even be big HPC. But if there is even a modest population of 1,000+ processor systems running Compute Cluster Server I would think at least one of them would be on the Top500 list.
Color me curious; I’m working on a feature piece right now where I’ll share what hard numbers I find out. If you are running, or know someone who is running, an HPC installation based on CCS, drop me an email.