In the case of a single job on a dual core then the job will get more than one core due to multi-threaded parts of the job, system tasks etc running on the second core. In this case the cpu is all accounted and the power of a core is known so the normalisation is correct.
In the HT case if you don't have one job per HT then the ones you do have actually appear run on faster CPUs so the normalisation is wrong.
If this really happens - sites don't run full all the time - then I expect one would see it in Steve's stats for ATLAS where he compares the actual events processed with the accounted time. This should change with the fullness of a cluster.
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Alex Martin
> Sent: 12 March 2012 09:52
> To: [log in to unmask]
> Subject: Re: CPU publishing (again)
>
> On Sunday 11 Mar 2012 19:20:12 John Gordon wrote:
> > I agree with Stephen. You should choose the number of jobs per node by
> > whatever criteria you want (HS06, Hammercloud, whatever) and whatever
> your
> > sweetspot number N is, run the HS06 benchmark with N jobs, divide the
> > total HS06 by N, publish N as the number of cores/cpu and HS06/N as the
> > benchmark/core. That way, as long as you configure the batch system to
> run
> > N jobs/node you will get the right installed capacity. I didn't
> understand
> > the comment about the accounting being wrong.
> >
> > John
> >
>
> Hi John,
> All the current accounting assumes the performance of the
> CPU
> performance per task is independent of the number of tasks running in
> total
> (which has probably never been the case). So if there are less than N
> jobs
> running they will tend to run faster than accounted for. Hyperthreading
> makes
> this much worst.
>
> Alex
>
> --
> ---------------------------------------------------------------------------
> ---
> |
> |
> | Dr. Alex Martin
> |
> | e-Mail: [log in to unmask] Queen Mary, University of London,
> |
> | Phone : +44-(0)20-7882-5033 Mile End Road,
> |
> | Fax : +44-(0)20-8981-9465 London, UK E1 4NS
> |
> |
> |
> ---------------------------------------------------------------------------
> ---
|