[Beowulf] Tyan's Beowulf in a box
Jim Lux
James.P.Lux at jpl.nasa.gov
Fri Jun 9 17:08:55 PDT 2006
At 03:10 PM 6/9/2006, Greg Lindahl wrote:
>On Fri, Jun 09, 2006 at 05:19:18PM -0400, Andrew Piskorski wrote:
>
> > These sorts of boxes might be more compelling to more people if they
> > included at least the option for HTX slots, PCI-Express slots, and/or
> > on-motherboard Inifiniband support.
>
>Those people can already buy blades or 1Us. I suspect that they (Tyan,
>Rocketcalc, etc) think their target market is people who want a small
>cluster, so gigE is good enough.
There are an enormous number of computational applications for which gigE
is "more than enough" for the cluster interconnect. Things like monte carlo
sims, or well defined computational electromagnetics problems that have
little need for between node comms. Example, I was just contemplating
running a model where I would need to calculate the electrical near field
at several million points covering a volume (a 20x20x20 meter volume
sampled at 10cm intervals), and it's trivially partitionable.
James Lux, P.E.
Spacecraft Radio Frequency Subsystems Group
Flight Communications Systems Section
Jet Propulsion Laboratory, Mail Stop 161-213
4800 Oak Grove Drive
Pasadena CA 91109
tel: (818)354-2075
fax: (818)393-6875
More information about the Beowulf
mailing list