Job Length Estimation and Performance in Back lling Schedulers

Back lling is a simple and e ective way of improv ing the utilization of space sharing schedulers Sim ple rst come rst served approaches are ine ective be cause large jobs can fragment the available resources Back lling schedulers address this problem by allowing jobs to move ahead in the queue provided that they will not delay subsequent jobs Previous research has shown that inaccurate esti mates of execution times can lead to better back lling schedules We characterize this e ect on several work loads and show that average slowdowns can be e ec tively reduced by systematically lengthening estimated execution times Further we show that the average job slowdown metric can be addressed directly by sorting jobs by increasing execution time Finally we modify our sorting scheduler to ensure that incoming jobs can be given hard guarantees The resulting scheduler guar antees to avoid starvation and performs signi cantly better than previous back lling schedulers