In order to reduce the energy waste of cloud data center and improve the Quality of Service(QoS), a adaptive scheduling policy of virtual machine(VM) service mode is proposed. Based on this policy, a reasonable customer charge scheme is developed for the cloud computing platform manager. According to the current number of system request nodes, the service rate and service mode of the system are dynamically adjusted in order that VMs can be managed by partitions. The service platform is modeled based on the M/M/c+k/d queueing system with partial servers synchronous vacation and working breakdowns, then the stationary distribution and performance indices of the system is obtained by using the iterative method. The numerical analysis demonstrates the influence of system parameter on the performance indices. The results show that the overflow probability of adding 12 VMs in zone II is 7.3% lower than that of adding 12 VMs in zone I. Meanwhile, the scheduling policy has a significant energy conservation effect. Furthermore, by constructing the benefit function, the customer charging strategy that modulates individual profit and social profit is derived.The establishment of this model will provide theory basis and tool for the optimization management and control of the cloud computing platform.