Hi
when I killed a large portion of WMS analysis jobs in Sheffield, the
netrwork load certainly decreased, the number of failures in production
and analysis pilots jobs (which were caused by timeout errors) decreased
as well. Howerever, killing jobs didn't help WMS survivors to increase
efficiency.
Elena
On Thu, 11 Jun 2009, Stephen Childs wrote:
> Graeme Stewart wrote:
>> The nominal AOD event size is 200kB, so a 2GB merged AOD has about 10k
>> events.
>>
>> The ganga jobs in hammercloud read about 10-15GB of data, usually, so
>> ~50-75k events.
>
> Thanks for the guidance.
>
> I currently have 223 of the ganga jobs running and they're still saturating
> the network. I'm inclined to kill off 50-100 of the more recent so that at
> least some useful work can be done by the survivors.
>
> Stephen
>
> --
> Dr. Stephen Childs,
> Research Fellow, EGEE Project, phone: +353-1-8961797
> Computer Architecture Group, email: Stephen.Childs @ cs.tcd.ie
> Trinity College Dublin, Ireland web: http://www.cs.tcd.ie/Stephen.Childs
>
____________________________________________________________________________
Dr Elena Korolkova
Email: [log in to unmask]
Tel.: +44 (0)114 2223553
Fax: +44 (0)114 2223555
Department of Physics and Astronomy
University of Sheffield
Sheffield, S3 7RH, United Kingdom
|