Dynamic Reliability-Optimised and Energy-Efficient Scheduling Algorithms in Heterogeneous Multi-core Systems

In the field of task scheduling algorithms for multi-core processors, energy consumption and reliability are always crucial topics. Some static scheduling algorithms attempt to reduce energy consumption while satisfying reliability goals by combining task replication techniques with DVFS technology. However, the static scheduling algorithm doesn’t consider the processing of redundant task replicas, resulting in waste of energy consumption. In this paper, we aim to improve the traditional static reliability-optimised and energy-efficient scheduling algorithm in heterogeneous multi-core systems, considering the dynamic processing of redundant task replicas. We propose a method for priority calculation, OC, and an algorithm for dynamic task replica deletion, DTRD, which we combine with the task replication algorithm RRES from existing research to obtain a new dynamic scheduling algorithm, D-OREES. To test the generality of DTRD algorithm, we also combined DTRD algorithm with the existing EFSRG algorithm as D-EFSRG algorithm. In the experimental part, the simulation scheduling results of FFT applications, GE applications and random applications show that the dynamic scheduling algorithm combined with the DTRD algorithm save more than 20% of the energy consumption compared with the original static algorithm while satisfying the same reliability goal, and the scheduling length is shorter. Among the four algorithms, the D-ORRES algorithm proposed in this paper has the best performance.