[Beowulf] Please help to setup Beowulf

Glen Beane Glen.Beane at jax.org
Fri Feb 20 10:37:05 EST 2009




On 2/20/09 10:22 AM, "Reuti" <reuti at staff.uni-marburg.de> wrote:

Am 20.02.2009 um 14:36 schrieb Prentice Bisbal:

> Bill Broadley wrote:
>> SGE does seem the default for
>> many small/medium clusters these days (and is the Rocks default)
>> but does make
>> some things strangely hard, usually with a work around though.  In
>> particular
>> I find the lack of a straight forward way to handle requesting
>> nodes and
>> processors per node strange.
>
> Me, too. Users are asking me for ppn=x functionality like Torque/PBS
> has, and I'm finding pretty complicated to implement on SGE.

Well, you have to setup nearly identical PEs like openmpi2, openmpi4
with a fixed allocation rule of 2 or 4 respectively.

OTOH: Once I looked into Torque and found, that with "nodes=2:ppn=2"
I got just one node with 4 slots in total of course. I don't know,
whether it can still happen to get such a distribution.


yes,  you certainly can get such a distribution with torque.  Moab and Maui can be configured to remap a nodes=X:ppn=Y to some other configuration to improve utilization,  but you can turn that feature off and then the user gets precisely the distribution that they request.  As far as I know work is being done to allow the behavior to be chosen at submit time.



--
Glen L. Beane
Software Engineer
The Jackson Laboratory
Phone (207) 288-6153


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.clustermonkey.net/pipermail/beowulf/attachments/20090220/c3614e58/attachment-0001.html>
-------------- next part --------------
_______________________________________________
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