Thermal and energy-aware utilisation management on MPSoC architectures

Abstract

High operating temperatures have been a major problem in embedded systems due to high throughput and compact designs required by modern applications. This paper introduces a novel strategy to subdue these high peak temperatures of MPSoC systems by incorporating task migration and task swapping, to eliminate hot spots and thermal gradients. The work further confirms that the proposed techniques help mitigate power consumption while maintaining a high throughput to satisfy the deadline requirements of the tasks in the system. A heuristic approach is presented that helps mitigate the hot spots and gradients specifically formed from the high-frequency execution of high-priority tasks. In addition, energy-aware task migration reduced the energy consumption in the system. Extensive experimental testing on actual hardware and simulation showed very plausible results to confirm the capability of the presented techniques to reduce peak temperatures along with reduced energy consumption in the system. The presented techniques performed better than many other standard and state-of-the-art published approaches in the literature.