forked from kjellp/SMRT-Analysis
-
Notifications
You must be signed in to change notification settings - Fork 0
The head node is very slow when more than 20 jobs are running
pb-jlandolin edited this page Mar 7, 2013
·
1 revision
The head node can becomes extremely slow when more than 20 jobs are running simultaneously.
The current distributed system can potentially use resources and create processes on the head node, for example, if the qsw.py script is used. Too many simultaneous running jobs may impact performance on the head node.
In practice, we have had no issues with 20 or more simultaneous jobs (resequencing with consensus). When running GATK, we did not see this issue with 80 or more simultaneous jobs.