I almost missed the release Compute Cluster Server (CCS) happened at the start of June; I wasn't expecting much to happen on the server side.
I'll confess that missed the point of CCS. I've seen the list on Top500.org topped by an IBM system with 128K processors and a speed rating of 280 Terraflops. Were we really getting into to the market for systems costing millions or tens of millions of dollars ?   

I seem to be turning into Hugh McLeod's bloggleganger. I responded to something he wrote about Microsoft needing a new big idea I said there was still plenty of mileage in the big ideas we had already:

  1. The power of the Personal Computer. Local processing is (usually) good.
  2. Make it easy for other people to use your product in their solution. This applies to OEMs who make PCs and devices, partners with solution for a specific market [both of these become your sales force] and end users.
  3. The power of Volume. Cheap hardware gives a huge market. If you sell Hundreds of Millions of copies of Windows, you can spend billions on developing it, and still sell it relatively cheaply, etc

So back to supercomputers. Only 10 of the top 500 have 256 processors or fewer - these are exotic beasts: not mainstream but low volume, difficult and impersonal, remote beasts. Otherwise they'd be ordinary computers.

I started to "get" CCS when I read a post on John Powers blog. Powers is CEO of Digipede, a Microsoft Gold Partner who produce distributed computing solutions, so he should get this stuff. Powers was responding to a member of the Linux community who had said that he
"wouldn’t be surprised if by 2007 Windows holds 2, maybe 3, whole spots on the top500 list."
Here's Powers' response:
"the Top500 list, by definition, rounds to approximately zero percent of the market for high performance computing.  Also, it represents the zero percent that Microsoft specifically identified as outside its target market... Microsoft is going after not the Top500, but the next 500,000....  Microsoft has been very open about this, saying they would bring HPC into the mainstream"

Ah !  High Performance Computing is broader than those exotic supercomputers which have power measured in Terraflops - thousands of times the power of a desktop PC, at a cost of millions of dollars. Is there demand for HPC solutions which are tens of times the power for thousands of dollars? That's what we're aiming for ... Volume.

Then I found the WinHEC presentation "Windows Server Compute Cluster Server and Desk-side Clusters" from the CCS home page and I can recommend it to anyone who wants an introduction to this area. Desk side clusters are the Local dimension The home page had a link to the inquirer - not normally a pro-Microsoft site. Their article "Microsoft sees HPC for non-geeks" makes my point about making it easy for people to "use your product in their solution"
Microsoft has a lot going for it. Rock-solid relationships with the IHV community mean that IBM, HP and Dell will be ready in August with pre-configured boxes and its undoubted nous at recruiting ISVs is also likely to pay dividends.

Tagged as