[Beowulf] Didn't we see this before?

[Jim Lux]

At 09:20 AM 6/24/2004 -0500, you wrote:
>Do we have any particular reason to think this foray will produce a 
>Windows implementation that will offer significant benefit to the science 
>and engineering developers over what we're doing?
>
>At the risk of being accused of taking things out of context, I 
>particularly liked this quote:
>
>"Parallel-computing clusters are increasingly being seen in the 
>enterprise," Microsoft product manager Dennis Oldroyd says. "It's been the 
>domain of academia and research. Now, with low-cost standardized hardware, 
>it's becoming less of a niche play."
>
>So: have we been relegated to niche status?  Somehow I thought what I'd 
>done with geodestic processing codes in the past was closer to 
>leading-edge than niche, but I guess, since I wasn't using Windows, I was 
>deluded.
>
><http://www.informationweek.com/story/showArticle.jhtml?articleID=22101604>
>
>-

A niche to Microsoft is a huge gaping hole to the rest of the world.

I like the statement about growing popularity of Linux on high performance 
clusters (hmmm. what's Linux replacing?)

It will all come down to support and integration... Will MS support the 
hardware interconnects (or, will the hardware vendors want to write MS 
drivers for their hardware)? Will MS provide decent hooks into their 
development environment (.NET, Visual Studio)?
How much of an ordeal will it be to take a useful code running on a Linux 
cluster and port it over to your shiny new Windows Server2003/HPC edition 
cluster? Is it just configure/make?  What will the licensing work like? If 
we change the hardware configuration, will WS2003/HPC need to "phone 
Redmond" to authorize?  Does WS2003 support NFS mount points? (Win2000 doesn't)

For "from scratch" development, it's probably not a real big deal which 
environment you develop for (assuming that they both "work"), and 
presumably, the OS isn't consuming the majority of the CPU cycles in either 
case.


James Lux, P.E.
Spacecraft Telecommunications Section
Jet Propulsion Laboratory, Mail Stop 161-213
4800 Oak Grove Drive
Pasadena CA 91109
tel: (818)354-2075
fax: (818)393-6875