COMFAST: A comparative framework for analysis of scheduling techniques in multi-core systems

Multi-core systems are increasingly being used for developing many real-time systems to meet the growing demands of high-speed computing and low-power dissipation. However, the delays due to the conflicts that arise when multiple cores access the shared system resources simultaneously encumber these systems to reach their peak efficiency. Several scheduling algorithms have been proposed to abate the issue of resource contention. However, comparing various scheduling techniques is a major challenge due to the different underlying assumptions and the lack of a common ground for analysis. In this paper, we propose a unified Comparative Framework for Analysis of Scheduling Techniques (COMFAST) in multi-core systems. The proposed framework supports the comparison based on a general data set and characterizes the performance in terms of the worst-case response times (WCRT) and acceptance ratios. For illustration, the paper presents the comparison of two of the recently proposed scheduling techniques, i.e., 1) Dynamic Task Aware Scheduling that uses complimentary tasks to mitigate contention, and 2) First Fit Allocation Based on Symmetric Analysis that approaches response time through the shared-resource-centric perspective and the core-centric-perspective. Our results show that Dynamic Task Aware Scheduling provides considerably reduced response times for a variety of resource access frequencies compared to First Fit Allocation. However, in terms of acceptance ratios, First Fit Allocation demonstrates a much better performance.