3E: Energy-Efficient Elastic Scheduling for Independent Tasks in Heterogeneous Computing Systems
Format of Original
Journal of Systems and Software
Original Item ID
Reducing energy consumption is a major design constraint for modern heterogeneous computing systems to minimize electricity cost, improve system reliability and protect environment. Conventional energy-efficient scheduling strategies developed on these systems do not sufficiently exploit the system elasticity and adaptability for maximum energy savings, and do not simultaneously take account of user expected finish time. In this paper, we develop a novel scheduling strategy named energy-efficient elastic (3E) scheduling for aperiodic, independent and non-real-time tasks with user expected finish times on DVFS-enabled heterogeneous computing systems. The 3E strategy adjusts processors’ supply voltages and frequencies according to the system workload, and makes trade-offs between energy consumption and user expected finish times. Compared with other energy-efficient strategies, 3E significantly improves the scheduling quality and effectively enhances the system elasticity.