But isn't that because there is a separate field for jobslots. It is a
function of the CE-queue, not a cpu.
John
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Ewan MacMahon
> Sent: 23 November 2009 12:19
> To: [log in to unmask]
> Subject: Re: Cores, CE_PHYSCPU, CE_SMPSIZE, CE_LOGCPU
>
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes [mailto:TB-
> > 2009/11/23 Rob Fay <[log in to unmask]>:
> > >>
> > > Out of curiosity, what if you don't run 1 job per thread? E.g. if
> you
> > > limit the box to, say, 12 job slots (for reasons of disk or
network
> > > contention perhaps), what do you publish then?
> >
> > Then you publish 12 logical CPUs, otherwise you scare the animals
> > (confuse the assumptions made about how logical CPU maps to "number
> of
> > allowed jobs".
> >
> Not according to the spec you don't - you publish 16. The PDF linked
> earlier explicitly declaims any advertising of job slots:
>
> "LogicalCPUs counts the number of computing units seen by the OS on
> the WNs. Sites typically configure one job slot per logical CPU, but
> some sites allow more than this (e.g. five jobs per four cores) to
> allow for the fact that jobs spend some time waiting for i/o. The
> 1.3 GLUE schema does not allow such an over allocation to be published
> explicitly."
>
> Ewan
--
Scanned by iCritical.
|