Re: [HTCondor-users] Minimize time to start job.

[Greg Thain <gthain@xxxxxxxxxxx>]

On 8/13/19 6:21 AM, don_vanchos wrote:
> Hello,
>
> I noticed that for the simplest vanilla jobs on my cluster, the time 
> difference `job_ad["JobStartDate"] - job_ad["QDate"]` is from 5 to 20 
> seconds. So quite a lot of time elapses between sending job to the 
> queue and starting the process.


Note that in HTCondor, the time to start the first job may involve a 
delay, because the schedd needs to have the negotiator select a machine 
for it to use. The NEGOTIATOR_CYCLE_DELAY can influence this time. 
However, once the schedd has a matching machine, it can re-use that 
machine for subsequent jobs that also match. The reason 
NEGOTIATOR_CYCLE_DELAY exists is if there is more than one schedd in a 
pool, the negotiator could assign a machine to the first schedd, and if 
the first schedd hasn't gotten around to claiming the machine, the 
machine will be idle, and could be re-assigned in a following negotiator 
cycle.

-greg