In this paper, we model a cloud server farm as a finite capacity queuing model where the rejection loss and impatient user loss are both considered. The effects of service rate and buffer size on the mean number of tasks in queue and system rejection probabilities are both demonstrated. The management center in our cloud model focuses on the optimal service rate and buffer size control. A heuristic algorithm is developed to maximize provider's profit and solve the conflict problem between system performances and cost savings. Simulation results show that our approach not only can maximize cloud provider's profit but also can satisfy the service level agreement (SLA).
展开▼
