A Lyapunov-based approach for differentiated service admission control and scheduling in cloud

Open cloud platform provides compute, storage and network resources as services to application developers. It is unrealistic to require all the third-party developers to develop stable applications. Hence admission control and reasonable scheduling of all the requests are very necessary for the platform stability. Based on the theory of Lyapunov drift and Lyapunov optimization, a differentiated service admission control model for open cloud platform is proposed in this paper. This model supports heterogeneous back-end Web cluster, and can distinguish between different types of requests. In the paper, we give a specific admission control and resource scheduling algorithm, which achieve an objective value only worse than the optimum by a constant factor. This property is achieved without any statistics of the arrival flows. Finally, we give a formal proof and some simulations, the result shows our algorithm runs well as we expected.