A-S algorithm: An optimal on-line real-time scheduling algorithm for uniform multiprocessors
Luo Xiaojian, Long Xiang · 2017
Scheduling periodic real-time tasks upon uniform multiprocessors is studied in this paper. We propose an optimal scheduling algorithm on uniform multiprocessors, named A-S algorithm, which achieves on-line scheduling as PCG[5] algorithm. PCG assigns task with largest local remaining time to fastest idle processor when rescheduling, which incurs great context switches between consecutive rescheduling. A-S keeps as many processors' assignments unchanged as possible at rescheduling points through a greed algorithm. The resulting A-S schedule can reduce preemptions and migrations dramatically (90.8% and 87.5% at most, respectively) when compared to those of PCG schedules.