[Beowulf] MS HPC... Oh dear...

[Robert G. Brown]

On Mon, 12 Jun 2006, Joe Landman wrote:

>> As I said, ROTFL.  That works fine for numb-nuts spending $500.  It
>> doesn't work that well for corporate or government decision makers
>> controlling the disposal of $500,000, where the question is whether it
>> buys (say) 2000 Linux nodes or 1000 Microsoft HPC nodes.  Somebody's
>
> Note to self:  Find out who the heck is selling Opteron servers for
> $250/node (see RGB's math above).  :)

Oops.  Why do you think I need computers?  Besides, what's a factor of
ten...

Most of the rest of what you said I don't strongly disagree with,
actually.  I just think that the introduction of WinXX clusters won't
decrease the number of Linux clusters, by in large, or make it
tremendously less likely that new ones get built.  It will largely make
inroads into new, fairly specific markets.  It will let certain classes
of users with certain needs set up "a cluster" with some marginal
benefit.  However, to succeed on the broader market front, it will have
to be truly cost effective, not just niche effective.  I suspect that
we'll see a market not unlike the server market emerge, with MS holding
a share.  However, the share will be much smaller than their server
market share is now, and that share isn't THAT big and is only growing
as MS and Linux split up the (ex)-Sun Lunch.  One day there won't be any
more Sun server operations left to split, and we'll see Lin and Win head
to head in the server market.  I predict a steady, inexorable shrinkage
of Win in favor of Lin at that point (just as now Lin gets roughly two
new operations to every one MS is getting).  But we'll see...

    rgb

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu