[Bioclusters] BLAST/ PBS / Grid Engine

[Ron Chen]

--- Chris Dagdigian <dag@sonsorol.org> wrote:
> that there is overhead involved in
> submitting a job, 
> having it be accepted by the scheduler and then have
> it dispatched for 
> execution, return the output etc. etc.
> 

In SGE, you can tune the qmaster so that the event
reporting peroid can be much less than 30 seconds.
I've never done it myself.

The way it is supposed to work is that when there are
no jobs submitted or finished, the scheduler is not
notified, so that the system load is not increased
that much.

(SGE is more event-driven than LSF)

> There is one other catch though that I remember from
> a project at Research Genetics-- I think that in
> GridEngine that if there is no available machine
then
> your 'qsub' will exit with some sort of error. 

qsub does not exit even when you have no available
machines, I think you are talking about qrsh, which
when all the machines are busy, it prints out an error
message and exit.

So if you want to wait, you can either run qrsh with
the "not now" option, or you use a script to loop till
the system finds a machine.

-Ron



__________________________________________________
Do You Yahoo!?
LAUNCH - Your Yahoo! Music Experience
http://launch.yahoo.com