[Condor-users] Local universe scheduling

[Stuart Anderson <anderson@xxxxxxxxxxxxxxxx>]

Is there any way to short-cut or otherwise minimize the time spent
by a schedd waiting for the negotiator when running local universe
jobs?

The problem I would like to solve is how to run a large DAG with individual
jobs that run for only a few seconds on a schedd machine that has 16 CPU-cores.
However, this schedd is also capable of submitting jobs to a much larger
Condor pool and does not have its own dedicated negotiator. The local universe
jobs appear to get blocked waiting for negotiation cycles for the larger
pool, including jobs from another schedd that typically runs standard and
vanilla universes jobs.

Since all the current state about a particular schedd is held by that
schedd is there a way to have it make local decisions about running
local (and presumably scheduler) universe jobs?

Any ideas?

Thanks.

-- 
Stuart Anderson  anderson@xxxxxxxxxxxxxxxx
http://www.ligo.caltech.edu/~anderson