Performance evaluation of real time scheduling algorithms for multiprocessor systems

Abstract

In multiple processor systems scheduling, providing real time service guarantee plays an interesting and challenging role. In the recent years of investigation many scheduling algorithms have been proposed, analysed and evaluated. In this study, we intend the performance evaluation of three of these multiprocessor scheduling algorithms namely global Earliest deadline first (EDF), Proportional fair (PF) and a newer optimal algorithm Largest Local Remaining Execution First (LLREF). We empirically investigate the ability and overhead of all three schedulers for randomly generated task sets. Our results verify that LLREF outperforms PF. It also shows that global EDF can't feasibly schedule the task sets as LLREF and PF even though it is outperformer than other two schedulers. Finally our investigation concludes that LLREF scheduler feasibly schedule for any random task sets and also improve the performance compare to P-fair and global EDF.