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

Re: [Condor-users] lightweight cluster/grid engine

I really just wanted something "lighter weight".    Lower cpu/mem usage, not so many forked processes per job, primary means of execution is the "shell" (aka condor_run), not a crafted "submission file".  No "suspend/resume" stuff.  Looks carefully at the "load average" before starting new jobs on a machine.  Small codebase.  Attempts to automatically profile commands.  100k entries in the queue is OK.  Fast turnaround time so you can break large jobs into smaller units and have them safely "distributed" around a cluster.

There aren't (to my knowledge) many open source alternatives to condor, but I'm aware that condor is really best for long-running jobs (especially the ability to checkpoint).


On Mon, May 23, 2011 at 1:30 PM, Alan De Smet <adesmet@xxxxxxxxxxx> wrote:
Erik Aronesty <erik@xxxxxxx> wrote:
> Yes.  I've had some trouble having a job launch anoher job using
> condor_submit with the queue.   ( This among other things. )

What's the end goal?  I'm guessing some sort of workflow
management.   Did DAGMan not map to your needs well?