If we report T/2, the sites which run a faster re-schedule will get the
'first' jobs. I suspect the smaller sites would tend to be favoured
under these conditions.
We could also agree on a global minimum wait time to return, for example
120 seconds. This would be the minimum returned. It would probably be
less false than 0 seconds and give all sites an equal chance of running
the 'first' job when there is nothing in the queue.
Tim
--
Tim Bell
CERN Div IT/FIO, CH-1211 Geneva 23
Phone: +41 22 76 72370
E-Mail: [log in to unmask]
Office: 31 1-008
|