关键词: 自适应功耗管理，作业调度，遗传算法，高性能计算

Abstract: The job scheduling system is a core component of high-performance computer,and the goal of job scheduling is to meet the performance requirements undo the premise, and get the lowest total power consumption of all tasks. We presented a job scheduling strategy based on adaptive power management, The strategy is based on genetic algorithm,and takes the performance/power ratio of the job queue as the scheduling factor. Comparing with the traditional job scheduling algorithms, it largely increases the system's energy efficiency with ensuring the resources utilization rate and the job throughput as well as decreasing the resources pieces and the pending jobs. The experiments show that this strategy can effectively improve productivity, and reduce energy consumption about 9% compared with traditional strategy.

Key words: Adaptive power management, Job scheduling, Uenetic algorithm, High-performance computing