Most of the time, ATLAS use around 150 slots. This is a very rough
estimation. I have seen something like 270 slots for a few hours
(http://wt2.slac.stanford.edu/images/1monthJobSnapshot.png). I would be
eager to try 400 slots.
I think another question is that once ATLAS batch machines are up and
running, do we want them to be in general queues? I guess Randy will
certain want to see that happen:-) If yes, how do we calculate the ATLAS
shares?
Wei Yang | [log in to unmask] | 650-926-3338(O)
Randy Melen - SLAC/SCS wrote:
> Stephen, I don't think those numbers are right. Of course this is 2 days later than your email so
> ...
>
> Anyway, right now when I look, I see the following:
>
> If I consider all Linux jobs and job slots, **including systems in dedicated BaBar queues**, I see
> this:
>
> there are 3743 jobs running in 4930 possible job slots, about 78% full
>
> But if I look only at the job slots available for general queue jobs, then I see
>
> there are 525 jobs running in 538 possible job slots, about 98% full
>
> Anything over about 95% full is "goodness" since job termination and rescheduling etc.
> probably takes a few percent.
>
> On Tue, 13 Feb 2007 14:35:07 +0100 (CET) "Stephen J. Gowdy" wrote:
>
>> Hi All,
>> I see there are very few busy CPUs (around 691 out of 5918) just
>> now at SLAC. Can we increase the number used for ATLAS production? I guess
>> I'm not sure how the single NFS server will deal with a higher load. Wei,
>> do you have an idea of how far it can be pushed?
>>
>> regards,
>>
>> Stephen.
>>
>> --
>> /------------------------------------+-------------------------\
>> |Stephen J. Gowdy, SLAC | CERN Office: 32-2-A22|
>> |http://www.slac.stanford.edu/~gowdy/ | CH-1211 Geneva 23 |
>> |http://calendar.yahoo.com/gowdy | Switzerland |
>> |EMail: [log in to unmask] | Tel: +41 22 767 5840 |
>> \------------------------------------+-------------------------/
|