Can you provide more info about the type of jobs that you are trying to run on EC2??
1) How did you submit the jobs to Grid Engine? Did you request a PE (Parallel Envorinment)?
2) What type of instances (ie. nodes) are you using? And I assume you have multiple CPUs per instance.
Lastly, you can run qhost to find the load of each instance/node, and "qhost -j" to find the jobs running on the nodes:
Grid Engine / Open Grid Scheduler
----- Original Message -----
From: liang cheng <liang.cheng_at_gmail.com>
Sent: Wednesday, October 5, 2011 2:22 AM
Subject: [StarCluster] how to adjust the load of each node
I'm a newbie to starcluster. I just managed to created a 2-node
cluster based on AWS EC2. So far it is running the assigned jobs.
However, I found the total time of running my jobs in this cluster is
way slower than running all of them on one machine.
I'm suspecting the low performace is becauseI'm not using the full
capacity of each node.
Can someone let me know how to check the nodes' CPU load percentage of
each node ? How to read the load information from "qstat" ? If the
threshold is low, how can I adjust it ? Can I ask one node to accept
multiple jobs simultaneously ? Now my observation is that each node is
only taking one job at one time.
StarCluster mailing list
Received on Thu Oct 06 2011 - 08:19:11 EDT