On 04/03/2012 03:26 PM, Sam Skipsey wrote:
>
> Are ATLAS envisaging a workload that would cause a non-trivial risk of
> a large number of jobs all actively consuming 2GB of swap simultaneously?
We could easily handle (say) a couple of those jumbo-jobs
on an 8 core node, and (say) six small ones. And if we had
all jumbo-jobs but they only went into swap for short periods
at random times, it wouldn't be so bad. But if we had all
jumbo-jobs in swap all the time, then I'm afraid things would
take until hell freezes over to come out.
That's about the length of it. We'd have to run tests to
go over the whole parameter space to be sure, and draw
some graphs. There might be a (rather dry) paper in it for
someone!
Steve
--
Steve Jones [log in to unmask]
System Administrator office: 220
High Energy Physics Division tel (int): 42334
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 2334
University of Liverpool http://www.liv.ac.uk/physics/hep/
|