论文部分内容阅读
由于功耗的严格约束,现代嵌入式计算终端必须采用科学的动态功耗管理策略.文中在对计算机系统的动态功耗管理(Dynamic Power Management,DPM)模型深入研究的基础上,采用改进的DPM随机决策模型,从理论上证明了DPM最优策略是确定性马尔可夫策略,这为简化DPM控制算法提供了理论依据.在实例研究中,比较了空闲时间长度服从负指数分布与Pareto分布两种情况,发现经典的空闲时间长度服从负指数分布的假设与实际情况偏差很大.Pareto分布很好解释DPM超时策略在实际应用中可以取得优良节能效果这一现象.
Due to the strict constraint of power consumption, modern embedded computing terminal must adopt the scientific dynamic power management strategy.Based on the deep research of the dynamic power management (DPM) model of computer system, this paper adopts the improved DPM Stochastic decision-making model, it proves theoretically that the DPM optimal strategy is a deterministic Markov strategy, which provides a theoretical basis for simplifying the DPM control algorithm.In the case study, the comparison of the length of free time subject to the negative exponential distribution and the Pareto distribution two It is found that the assumption that the length of classical free time obeys the negative exponential distribution is very different from the actual situation.Pareto distribution is a good explanation for the phenomenon that DPM timeout strategy can achieve good energy saving effect in practical application.