Abstrakt

Optimized energy aware scheduling to minimize makespan in distributed systems

Rajkumar K, Swaminathan P

In the present world of large distributed systems and energy shortages, energy efficiency has become mandatory. One way to achieve this is by proper scheduling of tasks in a system of connected and dissimilar machines to minimize the energy consumption. Also, Jobs have to be scheduled properly to different systems in order to achieve maximum utilization of machines. “Makespan” has been the standard optimization criteria used in scheduling algorithms. Makespan is the time elapsed until all jobs scheduled are completely processed. A scheduling algorithm must look to minimize the makespan while satisfying the precedence constraints between the tasks. Though the problem is NP-hard, many algorithms have been proposed. There are algorithms proposed to reduce the energy consumption for uni-machine systems and sequential tasks. Our algorithm tries to reduce the overall energy consumption of a schedule in parallel and distributed systems by minimizing the idle state times of machines and thereby, keeping the makespan as low as possible. Our algorithm is more practical and executes much faster compared to the previous works on energy-aware scheduling.