Clustered virtual machine allocation strategy in cloud computing based on new type of sleep-mode and performance optimization

With the constant increase in the number and the scale of cloud data centers, the energy consumption control in cloud computing is becoming increasingly apparent. By introducing a periodic sleep mode with double control of wake up threshold and sleep timer, we propose a clustered virtual machine (VM) allocation strategy. All the VMs in a cloud data center are divided into two modules: The VMs in Module I are always awake, while the VMs in Module II will switch between sleep state and awake state according to the workload of the cloud data center. By establishing a queueing model with double service rates and (N, T) policy asynchronous multiple vacations of partial servers, and using the method of a matrix geometric solution, we evaluate the performance of the clustered VM allocation strategy in terms of average latency of cloud requests and energy saving rate of the system. Theoretical analysis results and simulation results verify the effectiveness of the proposed clustered VM allocation strategy. By constructing a system cost function from the perspective of economics, introducing an chaotic mapping mechanism and a nonlinear decreasing inertia weight strategy to the Particle Swarm Optimization (PSO) algorithm, we optimize the strategy parameters to achieve a reasonable balance between the response performance and the energy efficiency of the system. © 2020, Jilin University Press. All right reserved.