An energy saving strategy based on M/M/c queuing model with preemptive priority and asynchronous working vacation

Abstract

In order to reduce the idle energy consumption of virtual machines (VMs), the strategy of asynchronous working vacation of partial servers is introduced. Considering that users have different priorities in practice, we introduce the preemptive priority strategy and give the high-priority users the preemptive priority to ensure its service quality. A virtual machine (VM) scheduling strategy that based on queueing theory is proposed. By establishing a three-dimensional continuous time Markov stochastic model, the expressions of the mean sojourn time of user requests and the default rate of system are obtained by using matrix-geometric solution method and Gauss-Seidel iterative method. Then we define the total energy consumption, the influence of system parameters on each performance indicator is analysed through numerical experiments. Finally, Nash equilibrium and social optimal behaviours of users are studied, and we obtain the optimal arrival rate of user requests by numerical experiments.