Research On Resource Demand Prediction And Placement Optimization In Cloud Computing

Cloud computing performance and energy consumption are mainly affected by cloud computing resource scheduling strategy.In order to improve the quality of cloud computing services and reduce the operation cost of cloud data center,resource scheduling strategy has become a hot topic in the field of cloud computing technology.At present,some achievements have been made in the research of resource scheduling strategies such as resource prediction and deployment in cloud computing,but there are still some key problems to be solved urgently,such as the inability to accurately predict resource requirements and the difficulty of efficient placement of virtual machine resources under heterogeneous physical machine conditions.Therefore,in view of the above problems,this dissertation studies from two aspects: resource forecasting and resource placement.(1)A resource prediction algorithm based on neural network combination model is proposed.In view of the fact that the existing single resource forecasting model can not accurately predict the resource demand of cloud computing,this dissertation proposes a resource forecasting algorithm based on neural network combination model,NNCM,in combination with the existing resource forecasting algorithm.The NNCM model combines the results of multiple resource prediction models through neural networks,and optimizes the weight through model training.The NNCM model utilizes the existing resource prediction model to realize the accurate prediction of cloud computing resource demand by effectively combining the predicted results of the model.(2)A multi-dimensional resource placement optimization algorithm for heterogeneous physical machines is proposed.To solve the problem that heterogeneous physical machines are difficult to place multi-dimensional resources efficiently,this dissertation proposes a multi-dimensional resource placement optimization algorithm for heterogeneous physical machines——MRPOAHPM.MRPOAHPM algorithm is based on genetic algorithm.It analyses the placement of heterogeneous physical machines and multi-dimensional resources,and designs the optimization model of resource placement.The MRPOAHPM algorithm designs the PMSSABOG algorithm for the selection of physical machine specifications in the placement process.In the placement process,the evaluation function of placement scheme and the characteristics of genetic algorithm are used to optimize placement scheme continuously to achieve efficient placement of virtual machine resources.MRPOAHPM algorithm can find a near optimal placement scheme for virtual machines in a short time.(3)Based on the above resource prediction and resource placement algorithms,a number of simulation experiments were carried out to verify the feasibility and effectiveness of the proposed method.Compared with the mainstream resource forecasting methods,NNCM model can accurately estimate the needs of users for a period of time in the future,and the forecasting accuracy is higher,which lays the foundation for subsequent resource placement.Compared with the traditional multi-dimensional resource placement algorithms with the same physical machine and heterogeneous physical machine conditions,the CPU and memory resource utilization of MRPOAHPM placement scheme is higher,which effectively improves the utilization efficiency of cloud computing resources,reduces the response time of resource requests,and improves the quality of cloud services.