Energy efficient scheduling for a stream of applications on an MPSoC under throughput constraints

With the increasing utilization of multiprocessors System-on-a-Chip (MPSoC), many research problems are arising as challenging. One main issue is to increase throughput. However, this usually results in an increase in power and energy consumption. Increased power consumption is a major contributor to the formation of hotspots in the system. Hotspot can compromise the reliability and lifetime of the system. Dynamic voltage and frequency scaling (DVFS) is a technique widely used to reduce power consumption. However, power-aware scheduling does not always guarantee thermal safety and hence thermal-aware techniques are developed. In this paper, we design, test and compare scheduling techniques for a stream of applications under the objectives of maximum speed, maximum throughput, reduced power consumption under deadline constraints or reduced hot spots under deadline constraints.