Research On Data Center Resource Forecasting And Allocation Method For Green Cloud Computing

In recent years,with the advancement of industry digitization and the rapid development of cloud computing technology,customers and their demand for cloud computing resources have also grown rapidly and by leaps and bounds.On the one hand,cloud service providers need to be able to respond to user requests quickly and in a timely manner,provide cloud computing users with high-quality services,and meet users' computing needs.But on the other hand,data centers that are growing in scale consume a lot of power,but the current resource utilization rate of data centers is still low,causing unnecessary waste of resources and environmental pollution.Therefore,how to improve resource utilization and reduce cloud data center energy consumption through effective cloud resource management technology under the premise of ensuring the quality of service(Qo S)of cloud computing services is an urgent problem to be solved.However,despite many related studies,the current resource utilization rate of cloud resource management methods is still low,resulting in energy waste.This paper focuses on improving resource utilization,optimizing cloud data center resource allocation,and realizing high-efficiency and energy-saving resource management.The main research contents are as follows:1.For resource management of a single cloud data center,a CPU usage prediction algorithm based on Deep Belief Network(DBN)and Particle Swarm Optimization(PSO)is proposed,which is called DP-CUPA.The purpose is By improving the prediction accuracy of the CPU usage in the future,it is convenient for the resource manager to better manage resources,reduce energy consumption,and reduce costs.The method consists of three main steps.First,preprocess and normalize the historical data of CPU usage.Then,the autoregressive model and gray model are used as the basic predictive models,and they are trained to provide additional input information to train the DBN.Finally,PSO is used to optimize DBN parameters and predict CPU usage through DBN.Through a large number of experiments on the real data set of Google cluster usage tracking,the accuracy of DP-CUPA was evaluated.2.Aiming at the problems of low resource utilization and high energy consumption in the process of resource management in multiple data centers,a resource allocation method for multi-cloud data centers based on request prediction is proposed.The resource allocation method guides the resource allocation strategy to generate a resource allocation plan based on the predicted result of the application request.The resource allocation algorithm first allocates the cloud data center that receives it to the application request according to the current data center CPU utilization,and allocates it to the existing VM in the data center or temporarily constructed according to the virtual machine allocation strategy VM,through the physical machine allocation strategy,the VM will be assigned to the appropriate physical machine.Resource allocation based on request prediction can not only reduce unnecessary virtual machines and physical machines idle running,but also improve resource utilization and reduce energy consumption.The algorithm guides the resource manager to formulate an allocation plan in advance by predicting application requests in advance.Corresponding simulation experiments prove the effectiveness of the algorithm proposed in this paper.