基于能量模型的可分负荷调度算法的研究

为了解决分布式计算系统能量消耗成本高的问题,在具有动态电压调节技术计算处理器的基础上,研究总线型网络环境中可分负荷的能量调度问题.根据能量与处理器速度的N次幂关系,在忽略网络延迟和给定运行时间的前提下,以最小化能量消耗为优化目标,建立可分负荷调度问题模型.采用非线性规划方法和Kuhn-Tucker条件,提出新的基于能量模型的负荷调度方案,并设计了相应的算法流程.对新方案和其他调度方案进行了对比和分析,结果显示新方案能耗率减少了10%～30%,验证了新方案在节能方面的有效性和优越性.

Research on divisible load scheduling algorithm based on energy model

In order to solve the problem of high energy consumption ratio in distributed computing systems, this paper bases on processors that are capable of dynamic voltage scaling, and studies energy-aware scheduling problems about divisible loads in bus networks. According to N-time power relations between energy and speed of a processor, and under the premise of ignoring network delay and given deadline time, this paper targets minimizing energy consumption, and builds a problem model about divisible loads scheduling. Then, this paper uses non-linear programming and Kuhn-Tucker conditions, proposes a new divisible loads scheduling scheme based on energy model, and designs its programming flow. At last, it compares this new scheduling scheme with other schemes by experiments, data show its energy consumption ratio decreases 10% to 30%, and its effectiveness and superiority on energy saving is proved.