Optimization analysis of cloud system with preemption and adaptive asynchronous semi-sleep mechanisms

Abstract In order to reduce the system energy consumption and improve the system profit while guaranteeing the user service quality in cloud environment, also considering the external interference factors and the problems of frequent state transitions and service inefficiency caused by synchronized vacations of servers, this paper proposed a virtual machine scheduling strategy and established an M/M/c queuing model with preemptive priority, negative customer and partial servers asynchronous working vacation. The steady state probability distribution of the system was obtained by using the theory of quasi birth and death process, and the stationary performance indexes of the system, such as average queue length and system default rate, were obtained. The numerical simulation was presented to demonstrate the effects of the parameters on the steady-state performance of the system under the condition of different asynchronous vacation threshold. The individual profit and the system benefit functions were established, respectively, the optimal parameter values of the maximum benefit were obtained. Finally, taking the service rate of two kinds of customers as the decision variable and minimizing the average sojourn time of two kinds of customers as the objective function, a bi-objective optimization model is developed and the NSGA-II algorithm is applied to find the optimal solution set of Pareto. Under the given arrival rate, the optimal service rate is sought to achieve optimization design and management of the system.