[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] structuring parallel java jobs

This is a second hand request for which I apologize, but for some
reason the user who's looking to do this doesn't want to post to the
list, maybe he's shy...

The user ahs a parrallel algorithm he wants to implement in java and
run on our condor cluster.  As I understand his requirnets he wants to

1) know that all jobs of a given set are running
2) know where each is running so they can use socket communication
3) pause or halt the job set if a member is suspended or preempted

As he describes it this isn't a master/worker paradigm, though I
suspect running a "master" job that all the "works" check in with
might solve some of this.

without going into my various sugegstions all of which seem a bit
kludgy, does anyone have a way of running htis type of job that they
love (or atleast like)