On Thu, Mar 1, 2012 at 8:36 PM, Simon Matthews
<[email protected]> wrote:
> I setup an SGE cluster using the classic mechanism for the job spooling.
> However, I think that I am seeing some problems with that approach.
>
> In a typical use of our cluster, we throw about 10000 jobs into the queue in
> the space of 15 minutes. These jobs run in the space of the next 3-4 hours.
>
> However, I think that I may have problems with performance of the cluster
> due to the number of jobs in the queue at any given time. The file
> <clustername>/spool/sge_job is currently 132MB.

Looks like you already are using BerkeleyDB spooling or else you
wouldn't be seeing the "sge_job" growing.

So what's your hardware config for the qmaster?

Rayson



>
> I have a very small number of machines in the cluster (less than 10).
>
> Is there a not-overly difficult way to convert a cluster to use bdb
> spooling? Is this a good idea?
>
> Simon
>
> _______________________________________________
> 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