[torquedev] nodes, procs, tpn and ncpus

"Mgr. Šimon Tóth" SimonT at mail.muni.cz
Fri Jun 11 12:15:06 MDT 2010


On 11.6.2010 19:16, Martin Siegert wrote:
> On Fri, Jun 11, 2010 at 01:26:18PM +1000, Christopher Samuel wrote:
>>
>>    -----BEGIN PGP SIGNED MESSAGE-----
>>    Hash: SHA1
>>    On 11/06/10 04:27, Martin Siegert wrote:
>>    > I am with Glen: nodes=X is just an abbreviation for
>>    > nodes=X:ppn=1 - it alwasy has been that way.
>>    First I've heard of it, I've never seen that behaviour
>>    from Torque and we always used that (with the nodect hack)
>>    to let people request X cores on clusters (up until procs
>>    was introduced).
> 
> But with the nodect hack nodes=N behaves exactly like nodes=X:ppn=1:
> In that case the meaning of nodes=X:ppn=y is: "give me X*y processors
> distributed over nodes such that on each node any multiple of y
> processors is allocated." Consequently nodes=X:ppn=1 is the same as
> nodes=X. The huge problem was that none of our users who specified
> nodes=X:ppn=1 understood this behaviour and expected the "EXACTNODE"
> behaviour.

Seriously don't blame them. Telling someone that by writing nodes=X he
does request X cpus and not nodes is like telling him that ordering
water actually means beer.

-- 
Mgr. Šimon Tóth

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3366 bytes
Desc: S/MIME Cryptographic Signature
Url : http://www.supercluster.org/pipermail/torquedev/attachments/20100611/db895f36/attachment.bin 


More information about the torquedev mailing list