Temperature Aware Energy Management for Real-Time Scheduling

Nikhil Gupta and Rabi Mahapatra
Texas A&M University


Abstract

Reducing energy consumption and temperature are important issues for current and future microprocessors. In this work, we present a best effort Temperature Aware Dynamic Voltage and frequency Scaling (TA-DVS) scheme for real-time task scheduling using run time slack management. We experimentally conclude that when the system utilization is within a certain limit, energy management alone can satisfy system temperature constraint. Our proposed scheme, TA-DVS, reduces temperature constraint violations for tasksets with unconstrained utilizations by 18.9% on the average, compared to existing approaches.