Modeling the Performance of Heterogeneous IaaS Cloud Centers

Performance modeling of cloud computing centers is a challenging task due to complexity, shared resources and large scale of such systems. Introducing heterogeneous resources and workloads make the analysis even more complicated. In this paper, we propose a layered stochastic performance model applicable to cloud centers with wide range of heterogeneous physical and virtual resources as well as task characteristics without sacrificing the scalability, granularity and simplicity. Under various parameters and configurations, key performance metrics such as task blocking probability and total delay incurred on user tasks are obtained. The results also reveal practical insights into capacity planning for cloud computing centers.