Implementation and empirical comparison of partitioning-based multi-core scheduling

Abstract

Recent theoretical studies have shown that partitioning-based scheduling has better real-time performance than other scheduling paradigms like global scheduling on multi-cores. Especially, a class of partitioning-based scheduling algorithms (called semi-partitioned scheduling), which allow to split a small number of tasks among different cores, offer very high resource utilization. The major concern about the semi-partitioned scheduling is that due to the task splitting, some tasks will migrate from one core to another at run time, which incurs higher context switch overhead. So one would suspect whether the extra overhead caused by task splitting would counteract the theoretical performance gain of semi-partitioned scheduling. In this work, we implement a semi-partitioned scheduler in the Linux operating system, and run experiments on an Intel Core-i7 4-cores machine to measure the real overhead in both partitioned scheduling and semi-partitioned scheduling. Then we integrate the measured overhead into the state-of-the-art partitioned scheduling and semi-partitioned scheduling algorithms, and conduct empirical comparisons of their realtime performance. Our results show that the extra overhead caused by task splitting in semi-partitioned scheduling is very low, and its effect on the system schedulability is very small.