I think Martin is talking about the SI value he uses to calculate the
power of the cluster or subcluster, not what we publish in the BDII.
John
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Burke, S (Stephen)
> Sent: 31 October 2008 16:57
> To: [log in to unmask]
> Subject: Re: Specint
>
> Testbed Support for GridPP member institutes
> > [mailto:[log in to unmask]] On Behalf Of Bly, MJ (Martin)
> said:
> > We use the per core SI2K summed over cores - and each generation of
> > systems contributes a scaled number appropriate to its
> > performance. The
> > numbers published here agree with what I calculate using that method
> > so...
>
> I'm not entirely sure I understand that ... as I understand it the
> tier-1 always publishes an SI00 of 1000 and uses the internal PBS
> scaling to adjust for the speed of each WN. How do you set that
scaling
> factor? I would assume that you must do it by a per-core rating
because
> otherwise your CPU time limits would be very wrong for single-core
jobs
> running on multi-core nodes, and the APEL accounting would also be
> wrong
> ...
>
> Stephen
> --
> Scanned by iCritical.
--
Scanned by iCritical.
|