[Beowulf] just to stir the pot a bit more<grin>

Jim Lux James.P.Lux at jpl.nasa.gov
Mon Nov 28 13:57:15 EST 2005


So, in my mailbox this morning was the latest issue of "Microwaves and RF" 
with an article:
"Cluster Computers speed EM simulation" describing how Sonnet Software has 
clusterized their EM analysis tools, using the user interface on client, 
number crunching on cluster backend model.
http://www.mwrf.com/Articles/ArticleID/11466/11466.html

this is a good example of a task that is easy to partition, since the 
calculations for each frequency point are essentially independent of the 
others, so once you've built the model, you can easily farm it out to all 
the processors.

AND, it's a capability that is exceedingly useful to RF designers, who are 
always trading off wall clock time against the number of frequencies you've 
analyzed the circuit at, hoping you didn't miss some horrible resonance or 
interaction.

They use LSF from PlatformComputing http://www.platform.com/ as their 
cluster/grid/NoW manager


here's Sonnet's page
http://www.sonnetsoftware.com/products/em/ef_emcluster.asp

the pricing model is a bit punitive, IMHO, since they require a separate 
license of the top of the line software (Sonnet Pro) for EACH node (they 
don't quote a price on the website, but the tier BELOW the Pro (Level 3 
Gold) is $12K)..

Seems that all the stuff in the article about how cheap each node can be is 
sort of pointless, when you're going to be installing software that costs, 
easily, 4x the hardware cost.

No data on what the LSF component from Platform might cost (the Platform 
website is very flashy, but relatively content free, unless you like 
reading press releases)


However, it also appears that the backend could be Linux and the front end 
client could be Windows, since Sonnet runs on all of them.  (or, just to 
get rgb cranked up... the single client could be Linux, and the 
computational back end could some horrible mix of Apple, Unix, and Windows 
machines, perhaps of different versions!)



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


_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list