A reliability-aware and energy-aware task scheduling algorithm for heterogeneous multi-core systems
Dongyang Liu, Junhua Chen, Xueda Huang, Haojun Hong · 2024
Heterogeneous multi-core systems are widely used in high-performance computing applications. Reliability and low power consumption are important performance indicators of existing heterogeneous multi-core systems. The task scheduling algorithm with these two joint indicators as the optimization goal has important research significance. This paper designs a task scheduling algorithm for reliability and energy consumption awareness of heterogeneous multi-core systems. The algorithm is based on periodic real-time tasks with dependencies, and uses the redundancy of heterogeneous multi-cores to propose primary and secondary version backup technology for fault tolerance. The backup technology unifies the main task and the backup task and combines them with the DAG task model to construct a new task model. Reduce power consumption by introducing DVFS and DPM technology. DPM technology is mainly used for processor idle time, and consider using DVFS when the usage conditions of DPM are not met. In this paper, by comparing and evaluating randomly generated task models, the experimental results show that the algorithm has good performance in maintaining system reliability and reducing system power consumption.