Change to the Scheduler Policy

Posted by carlo on 18 Jan, 2018

Dear contestants,

given the numerous requests, we have updated the scheduling policy of the jobs in the testbed.

Instead of a FIFO policy (i.e., jobs are executed based on the date/time in which they were queued), we will schedule experiments in a round-robin fashion. Once the job of one team has completed, we will execute the next pending job for the next team number (if any) and iterate through all other competing teams.

Please note that, (i) once the testbed starts scheduling jobs, a random team is selected to execute first, and that (ii) if there is no job pending for a given team, the scheduler will first iterate through all other teams before scheduling a newly pending job.

This should increase fairness in the number of experiments executed per team in a given time window.