[Beowulf] How Can Microsoft's HPC Server Succeed?

[Bogdan Costescu]

Sorry to divert a bit the thread towards its initial subject and away 
from the security issues currently discussed...

I've just seen a presentation from a University (which shall remain 
unnamed) which partnered with Microsoft for... well, HPC. The reasons 
for using Windows were more or less the same that have been mentioned 
in this thread, so I won't repeat them. To note is that they weren't 
using Windows exclusively, but only on a part of the cluster, the rest 
running Linux.

Towards the beginning of the presentation there was a mention of a MPI 
latency benchmark showing 2.something microseconds over their IB 
(unknown make and speed) in mainboards using latest generation Intel 
CPUs with Microsoft's MPI libs, which seemed like a decent performance 
and got me pretty excited. But then I changed my mind when I started 
to hear what a great feature it is to have several nodes booting and 
installing the OS in the same 50 minutes (yes, minutes!) that a single 
node takes, due to a wonderful feature called multicast. And then 
things turned really strange after a statement saying that in Linux it 
takes several minutes to start a parallel job while in Windows only 
about 10 seconds. Then I started wondering: were those 2.something 
microseconds a measure of the same latency that I know of ?

I can't say for sure that this was part of some Microsoft strategy and 
not a PR effort gone bad, but I'm strongly enclined towards the first 
which leads me to believe that the answer to the question in the 
subject is: by disinforming people. Yes, there are probably many CEOs 
of SMBs, who don't know/care much about technical details and don't 
have a clue about Linux HPC, who are going to be impressed by such 
statements. And when you can run HPL from Excel by modifying in a cell 
one of the parameters and getting the results back in that table, 
results from which you can quickly generate a graphic and say "whew, 
I should be in Top500", who can say that clustering is hard and 
user-unfriendly ?

I'm all for healthy competition in this area, especially as I think 
that HPC didn't evolve significantly in the past few years. But such 
aproaches are far from healthy... well, at least for my definition of 
healthy competition. ;-)

-- 
Bogdan Costescu

IWR, University of Heidelberg, INF 368, D-69120 Heidelberg, Germany
Phone: +49 6221 54 8869/8240, Fax: +49 6221 54 8868/8850
E-mail: bogdan.costescu at iwr.uni-heidelberg.de