摘要
提出了一种基于马尔可夫切换状态空间控制模型的多媒体服务器集群系统能耗最优控制方法.通过建立多媒体服务器集群的随机控制模型,将能耗最优控制描述为一个带约束的优化问题.结合拉格朗日乘子法和性能势理论,提出了一种在线策略迭代算法.该优化算法通过样本轨道在线寻找最优控制策略,寻找过程不需要精确的系统参数信息.仿真实验证明了该算法的有效性.
An energy conservation optimal control approach is proposed for multimedia server cluster systems based on Markov switching state space control model. First the stochastic control model is introduced for multimedia server cluster systems. Under this model, the energy conservation control problem is formulated as a constrained optimization problem. Based on Lagrange multipliers method and performance potential theory, an online policy iterative algorithm is proposed. The algorithm solves the optimal control policy on-line by sample path and the solving procedure does not need any accurate system parameter information. Simulation experiments demonstrate the effectiveness of the proposed approach.
出处
《信息与控制》
CSCD
北大核心
2013年第1期125-131,共7页
Information and Control
基金
国家自然科学基金资助项目(61074033)
教育部博士点基金资助项目(20093402110019)
中央高校基本科研业务费专项资金资助项目(WK2100100004)
关键词
动态能耗管理
马尔可夫决策过程
在线优化
性能势
dynamic energy conservation management
Markov decision process
online optimization
performance potential
