Graeme Stewart wrote:
> On Tue, Apr 21, 2009 at 5:05 PM, Sam Skipsey <[log in to unmask]> wrote:
>> 2009/4/21 Graeme Stewart <[log in to unmask]>:
>>
>>> I have one final question, which is about fairshare, does anyone
>>> understand the difference between DEDICATEDPS and DEDICATEDPES in
>>> maui? I want to make sure that fairshares are calculated based on
>>> wallclock times, not on cpu consumption times. The maui manual
>>> describes FSPOLICY (http://tinyurl.com/cjgpba), but doesn't really
>>> explain which setting does what.
>>>
>> I believe that DEDICATEDPS scales according to processor-seconds (that
>> is number of processors * time), and DEDICATEDPES scales according to
>> "effective processors"*seconds, and is designed to cope with
>> situations where, say, you're using 4 of the 8 cores on a machine, but
>> all of the memory. DEDICATEDPES would assign you an "effective
>> processors" of 8, not 4, since you're using more memory than you
>> should be (in fact, you're using all the memory in the box, so you're
>> effectively using 8 processors).
>
> Yes, this I understand. But if you have a job that user 1 core and
> sleeps for an hour does it get charged for its hour or not? And does
> the setting of DEDICATEDPES/DEDICATEDPS affect this?
>
Yes, it does get charged for its hour, and no, the setting of DEDICATEDPS or
DEDICATEDPES doesn't change that. DEDICATED usage tracks what the scheduler has
dedicated to the job, not what the job has consumed, so it is (essentially)
walltime. A 1 core job that sleeps for an hour will have a dedicated fairshare
usage of 1 proc-hour. A 4 processor job that runs four lots of sleep for an hour
will have a dedicated fairshare usage of 4 proc-hours.
I believe the alternative is UTILIZEDPS, which tracks actual usage of processor
time (although the maui manual is typically vague on this).
Rob
--
Robert Fay [log in to unmask]
System Administrator office: 220
High Energy Physics Division tel (int): 43396
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 3396
University of Liverpool http://www.liv.ac.uk/physics/hep/
|