Sounds like the NFS ops your LSF cluster generate may simply be overloading your 960 ... Have you tried any filers other than the 960?

The 980 filer has SPEC sfs numbers more than 40% faster than the 960.

If you really need high-performance try a filer even faster than a 980. On Spec.org you'll find the Spec numbers for the Titan filer, which delivers more than twice the spec. ops of a 960 filer and is almost 40% faster than a 980. That would probably help handle any LSF cluster load for a while.

http://www.spec.org/sfs97r1/results/sfs97r1.html Published Spec results - single head 960 25135 980 36063 Titan 50858

Published Cluster Spec. results 960c 48143 980c 68139 Titanc 101571 http://www.marketwire.com/mw/release_html_b1?release_id=75166

If you don't get a faster filer, the other option is to break-up the jobs into smaller parts.

Let me know what you end up doing ... could be useful and interesting.

Cheers Frank Patterson "Everything should be made as simple as possible, but not simpler." Albert Einstein

-----Original Message----- From: cisco joe [mailto:ciscojoe305@yahoo.com] Sent: Tuesday, 30 November 2004 2:45 AM To: toasters@mathworks.com Subject: Simulation jobs peaking filer CPU & CP

All,

We have some issues with LSF simulation jobs writing to filers(FAS960) ..

Filer's CPU peaking to 100% and NFS goes beyond 80K and CP back to back happening when users executes LSF regress simulations jobs...These jobs has heavy writes to filer and even read also high..

Because of this huge writes filer is going to hung state where simple ls or df from NFS is hanging ..Even telnet or rsh to filer will take long time ...

If anybody has any workaround pls let me know

Any suggestions is highly helpful

Thanks Joe

_________________________________________________________________ Don’t just search. Find. Check out the new MSN Search! http://search.msn.click-url.com/go/onm00200636ave/direct/01/