Prediction And Optimization Of Performance And Cost In Cloud Computing Virtual Machine Scheduling

As an essential feature of cloud computing,dynamic scalability enables the cloud system to dynamically expand or shrink resources according to user needs at runtime,which is the key to reducing the cost of cloud service providers and improving the efficiency of user interaction.At present,there is still no effective method to quantitatively measure the performance and cost of cloud computing systems,and complete the optimization of the overall configuration parameters of the system,so that effectively predicting and optimizing the cost and performance of cloud computing platforms has become one of the key research challenges in the field of cloud computing.In this paper,in order to quantitatively predict the cost and performance of cloud computing platforms,we propose two types of cloud computing resource analysis models,one considering the hot/cold startup and hot/cold shutdown of virtual machines(VMs),and the other considering the power consumption of VMs.In the first resource analysis model,multiple divisions of VM states are considered,which are divided into hot/cold startup and hot/cold shutdown.At the same time,based on the M/ M/ N/ ? queuing model,the stable probability distribution of the system is calculated through the quasi-birth and death process(QBD)and matrix geometric solution,which is used to analyze the performance and cost of the cloud computing platform and obtain accurate system indicators.The second resource analysis model considers dynamic power adjustment.It uses a queuing model and analyzes the VM measurement indicators at different power levels based on the stable probability distribution.The indicators include elasticity indicators,cost indicators,performance indicators,cost performance and so on.In addition,in order to optimize the overall configuration parameters of the system,different multi-objective optimization models are established for different resource analysis models,and then the optimal stopping algorithm and the proposed cost-performance optimization algorithm are used to obtain the optimal configurations.Among them,the first model configuration includes the number of hot startup VMs,the system service rate,the hot/cold startup rate of VMs,and the hot/cold shutdown rate.The second model configuration includes system service rates of different power levels,the number of VMs with the lowest power consumption level,and the transition probability among power consumption levels,etc.Finally,for system performance and cost optimization,through the analysis of experimental data,the single parameter optimization method and the overall parameter optimization method of the system are obtained.At the same time,the comparison with existing optimization methods proves the superiority of our proposed cost-performance optimization method.