[Zaphod-Users] job seduling
John Dorelli
john.dorelli at unh.edu
Fri Oct 7 10:22:58 EDT 2005
Hi everyone,
We are currently reorganizing the Zaphod job scheduling system.
Previously, Zaphod nodes were organized into partitions; and one
requested a particular partition in one's batch script as follows:
#PBS -W x=PARTITION:myrinet
We will be replacing the partitions with multiple queues which will be
mapped to different subsets of nodes. In the meantime, one should
request different sets of nodes by referring to node attributes (eth,
myri or storage). For example:
#PBS -l nodes=32:ppn=2:myri
will give you 32 myrinet nodes.
Apologies for any inconvenience; we should have a new and improved
scheduling system soon (hopefully today).
John
--
Dr. John C. Dorelli
UNH-EOS Space Science Center
245E Morse Hall, 39 College Road
Durham, NH 03824-3525
Office: (603)862-0613
Cell: (240)418-2582
john.dorelli at unh.edu
More information about the Zaphod-Users
mailing list