[mvapich-discuss] RHEL6.5 and MVAPICH2

Mehmet Belgin mehmet.belgin at oit.gatech.edu
Fri Aug 29 17:12:45 EDT 2014


Glenn, this is very interesting. We have not patched our kernel (currently using 2.6.32-279.5.2.el6.x86_64).  I will run some STEAM tests to see how it looks on our system.

Thank you very much sharing this information!

Mehmet


=========================================
Mehmet Belgin, Ph.D. (mehmet.belgin at oit.gatech.edu)
Scientific Computing Consultant | OIT - Academic and Research Technologies
Georgia Institute of Technology
258 4th Street NW, Rich Building, #326 
Atlanta, GA  30332-0700
Office: (404) 385-0665




On Aug 29, 2014, at 4:33 PM, Glenn K. Lockwood <glennklockwood at gmail.com> wrote:

> Mehmet,
> 
> Did you patch your kernel after upgrading?  There was a severe performance regression in EL 6.5 related to the Linux thread scheduler:
> 
> https://bugs.centos.org/view.php?id=6949
> 
> Glenn
> 
> 
> On Fri, Aug 29, 2014 at 12:15 PM, Mehmet Belgin <mehmet.belgin at oit.gatech.edu> wrote:
> Hi Everyone,
> 
> We upgraded from RHEL 6.3 to 6.5 recently and started noticing some performance fluctuations and increased level of sensitivity to core affinity (i.e. more gap between no affinity and affinity). Maybe this is selective bias, but we feel like mvapich2 (2.0rc1 build) performance characteristics have changed (not necessarily degraded, but seeing more inconsistency between runs).
> 
> Just pinging the list in case others might have observed this as well. Any feedback on this will be very much appreciated!
> 
> Thanks,
> -Mehmet (Georgia Tech)
> 
> 
> 
> 
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20140829/9155b252/attachment.html>


More information about the mvapich-discuss mailing list