[Biojava-l] Job launching system

[Keith James]

Sorry for mucking up the thread. I can't seem to follow up the post on
the list as the list server seems to have stopped sending me mail...

This message combines both replies:

>>>>> "Brian" == Brian Gilman <gilmanb@genome.wi.mit.edu> writes:

    Brian> Hello all, I'd be interested in this system as we have just
    Brian> put in a system to do this into omnigene. If you go to our
    Brian> cvs repo you can find all the SOAP classes and EJB's in
    Brian> edu.mit.wi.omnigene.webservice.analysis and
    Brian> edu.mit.wi.omnigene.analysis.

I've just had a browse. Omnigene is on my list of things to install -
I'm particularly interested in Tequila (why is that phrase so easy to
say?) and the middleware.

There was a cvs upheaval at the end of last week, but there's code at
cvs.sanger.ac.uk under Pathogen/psu/jobcontrol and
Pathogen/psu/genlib/java. I'm sure there's a good deal of refinement
needed (and more/updated docs).

    Brian> 	I have a few questions though...Have you written a JNI
    Brian> layer that interacts with LSF?? Or is this a command line
    Brian> program that does a runtime.exec() ??

I wrote a JNI layer which uses the LSBLIB C library. It only
implements a subset of the full LSBLIB API, but it's enough to submit
jobs and get back an LSF job ID and to query the system by job ID, job
name, queue and user. There are different implementations of the job
launcher interface - one uses LSF via JNI and the other uses
runtime.exec().

>>>>> And Mark writes

    >> Is there overlap here with J2EE? Not entirely sure what you
    >> mean by batch queues but I assume it would be along the lines
    >> of submitting jobs to a blast farm or similar.

I really would have liked to implement a server for this (and still
might), but given a limited time to get something working and the fact
that I had to learn SQL, JDBC and C during the process, I've left that
to one side for now. The persistence stuff is all behind DAO
interfaces, so if I've understood things correctly it should be
straightforward to set up.

The batch queues in our case are LSF queues for our compute clusters.

Keith

-- 

-= Keith James - kdj@sanger.ac.uk - http://www.sanger.ac.uk/Users/kdj =-
Pathogen Sequencing Unit, Wellcome Trust Sanger Institute, Cambridge, UK