Windows HPC

[Joachim Worringen]

Mark Hahn:
> oddly, this reads *exactly* like a msft press release.

Yes, indeed - this was my feeling, too.

> needless to say, the TCO/ease/managability claims are highly disputable.

Absolutely.

Nevertheless, it is not true if people claim "HPC on Windows is not feasible". 
As Paul said, there is interest to do this, and it probably works to some 
degree. Back at the university, we developed (and my colleagues still 
develop) an MPICH derivate which also runs on Windows 
(htttp://www.mp-mpich.de). From the downloads I can judge that there is a 
significant demand to do MPI on Windows platforms. Which isn't bad at all, I 
guess, because people get in touch with HPC this way which they might not if 
they had to switch to another OS in advance. As Paul stated, this broadens 
base for HPC usage. This is good, similar to the many beowulf-style 
installations which broaden the base for parallel processing, which in 
earlier days was reserved to relatively few centralized computing centres.

What I don't like about Paul's statement is the claim that Windows is the 
*better* solution for HPC. I.e. the claim "anyone can run Windows" - which is 
not true if it comes to set up a complex system. For an example, take a look 
at http://support.microsoft.com/default.aspx?scid=KB;EN-US;Q259267& - that 
looks so easy and straight-forward, I'll have my mother do it...

 Joachim

-- 
Joachim Worringen - NEC C&C research lab St.Augustin
fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de