Policy Of Energy Management For Cloud Computing Platform

High energy consumption is a major problem in the cloud computing system. Moreover, withthe growing scale of cloud computing in recent years, the energy consumption has also becomeeven more serious. This thesis focuses on the dynamic power management and resource schedulingto achieve the goal of energy saving for idle and extravagant energy consumption of cloudcomputing system. The main contributions in this thesis are described as follows:(1) The current situation of energy-saving technology is summarized. And then the relatedconcepts of cloud computing and dynamic power management strategy and resource schedulingalgorithm in cloud environmentnal are summarized.(2) An energy consumption estimation model based on different states is proposed in this thesis,which takes full account of the resource energy consumption in different states of sleep, idle,conversion and work, and this model is verified by a multifunction metering socket.(3) An adaptive idle time predictive policy based on exponential average algorithm is proposedin this thesis. This method imports the adaptive regulatory factors to adjust the influence ofpredicted idle time from historical idle time. Besides, this method combines with the idea ofsegmented sliding window. The idle time within the sliding window is divided into three categoriesas long, medium and short. This method takes the average idle time of the maximum number ofeach type as the actual idle time value of the exponential averaging algorithm. Through the forecastof next period of idle time, the policy decides whether to switch the state of the physical host toreduce idle power consumption. The experimental results show that the strategy of this thesis has ahigh rate of prediction accuracy and a little system response and a low energy consumption.(4) An energy-efficient resource scheduling algorithm in cloud environment is proposed in thisthesis for luxury energy consumption. First of all, resource scheduling in cloud environment ismodeled. Then, a minimum energy consumption resource scheduling algorithm based on improvedMin-Min algorithm Min-Energy is proposed, which reaches the target of minimizing energyconsumption of every task in a cloud environment based on meeting the QoS demand of cloud tasks.The tasks are sorted in accordance with the priority firstly, then the algorithm estimates the totalenergy consumption of each task in various resources and select the resource correspondingminimum energy consumption of each task for scheduling. The simulation on CloudSim platformresults show that the Min-Energy algorithm has better performance in the completion time and energy consumption, which achieves the purpose of energy saving.