Department of Computer Science at UH

University of Houston

Department of Computer Science

In Partial Fulfillment of the Requirements for the Degree of
Master of Science

Fang Liu

Will defend her thesis


An Energy Aware Management Technique for Real-time Multi-core Systems

Abstract

In this work, we propose Dynamic Migration Heuristic method integrating dynamic voltage scaling (DVS), dynamic power management (DPM) and task migration on multi-core real-time systems which can feasibly balance the task load and reduce energy consumption during execution to achieve energy efficiency. Meanwhile, a Voltage Scaling Based Dynamic Core Scaling approach is presented for reducing leakage power consumption under low task load conditions. The framework which is used for the proposed methods is composed of a partitioner, a local EDF scheduler, a power aware manager, a dynamic migration module and a dynamic core scaling module. The primary unit is the energy aware manager. The energy aware manager controls the power consumption and voltage scaling by the feedback of the dynamic migration and the dynamic core scaling modules. Simulation results show that our Dynamic Migration Heuristic can produce energy savings of about 3 percent as compared with the previous works. However, it greatly reduces the cost of task migration among the multi-core processors. The results also show that Voltage Scaling Based Dynamic Core Scaling can achieve up to 10 percent energy savings under low load conditions as compared with the previous works.

 

Date: Wednesday, July 17, 2013
Time: 1:30 PM
Place: PGH 563

Faculty, students, and the general public are invited.
Advisor: Prof. Albert M.K. Cheng