[GE users] Array Task Interleave

fx d.love at liverpool.ac.uk
Mon Apr 19 15:03:58 BST 2010


    [ The following text is in the "utf-8" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

reuti <reuti at staff.uni-marburg.de> writes:

> There is no statement I think, what will be scheduled first when you
> have A.5-A.100 waiting, and B.1-B.5 are eligable to run too as A.1-A.5
> finished. It depends also on the size of the cluster, what can run at
> the same time.

Sorry, I'm obviously not awake.

> And you can also attach an urgency to the B jobs by requesting a complex.

Concretely, we have:

  # qconf -sc | egrep 'high|low'
  high                hi          BOOL        ==      YES         NO         0        1000
  low                 lo          BOOL        ==      YES         NO         0        -1000
  
plus this in sge_request for default priority:

  # Sample #3 assigns a low default submit priority. For higher 
  # prior jobs the priority can be increased at jobs submission 
  # time by uing "-p <higher_priority>". The value of 
  # <higher_priority> may not exeed 0 because priorities above
  # this value can only be set by managers. 
  #
  -p -100

Are there good recommendations for values of the priority and urgency to
use like this, along with the various weights, that people find work
well for scheduling in typical situations (such as medium-sized cluster,
low throughput, ~10 active users, fair-share, for instance)?  I don't
remember seeing any discussion like that.

-- 
Dave Love
?E-Science?, Computing Services Department, University of Liverpool
AKA fx at gnu.org

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=254014

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list