Energy consumption on heterogeneous computing platforms

Abstract

In the recent past, heterogeneous computing has gained popularity over the traditional homogeneous computing. Moreover, the upcoming processors are typically equipped with dynamic voltage and frequency scaling options to reduce energy consumption as per the application requirements. Scheduling plays a vital role for efficient parallel computing systems, and in this work, the relationship among heterogeneity, make-span and energy consumption has been explored through simulations for heterogeneous processing systems. Heterogeneity is found to adversely affect the total energy consumption in comparison to the extent of makespan improvement.