Am 30.08.2013 um 19:57 schrieb Jennewein, Douglas M:

> We are running SGE 6.1u4 with Rocks 5.0 and are looking at implementing a 
> user chargeback model based on CPU-hour utilization.
>  
> We’d like to assign to a user or department a fixed number of CPU-hours, 
> ensure CPU-hours are available when jobs are submitted, and decrement the 
> available CPU-hours as jobs execute.
>  
> As I understand it, qacct will provide CPU-hour utilization for completed 
> jobs, and CPU-hour utilization for running jobs can be computed from qstat 
> output.
>  
> What other (better?) mechanisms exist in SGE to support something like this? 

There is nothing built in to charge any user/department directly. You can only 
implement it outside of SGE.

But it would be good to avoid that some nodes are idle because group A used up 
all their granted CPU time and waiting jobs, while group B has nothing to do 
right now.

If you are looking more in granting any group of users a certain amount of CPU 
time of the cluster over the last 30 days or so, you can look into the 
sharetree policy. SGE will try schedule the jobs in such a way, that the set up 
percentage for a user or department will be met over time.

http://docs.oracle.com/cd/E19080-01/n1.grid.eng6/817-5677/i1004072/index.html

-- Reuti

> Doug Jennewein
> Research Computing Manager
> Information Technology Services
> The University of South Dakota
>  
>  
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to