[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Cluster Job latency


I'm running some tests on condor 6.8.4 (X86_64-LINUX_RHEL3). When 
submitting a cluster job, the list of jobs appears as Running (I submit 
a job per node), but it seems that they are submitted every 2 seconds. 
At least the difference in RUN_TIME value between 2 consecutive jobs is 
usually 2s.

Does this correspond to a physical limit or can we reduce this value ?

Schedduling a task on 512 nodes takes about 17mn which is usually less 
than the computing time with that many nodes, so it limits the 
scalability of the application we use.

Matthieu Cargnelli
EADS Innovation Works
Campus Engineering
18, Rue Marius Terce
BP 13050
31025 Toulouse Cedex 03
Tel : +33 5 61 16 88 36
Fax : +33 5 61 16 88 05