Cost-Efficient Provisioning Strategy for Multiple Services in Distributed Clouds

Cloud platforms offer computing, storage and other related resources to cloud consumers in the form of Virtual Machines (VMs), and allow VMs scaling according to the workload characteristic. Specially, with cloud computing, service providers need no longer to maintain a large number of expensive physical machines, which can significantly reduce the cost. However, it is still a challenge for service providers to purchase the optimal number of VMs from distributed clouds due to the uncertainty of the service demands and the operational cost. To address this problem, in this paper, a Cost-efficient Provisioning strategy for Multiple concurrent Services (CPMS) in distributed clouds is proposed by formulating and solving a two-stage stochastic programming model. The objective of this model is to minimize the resource cost of purchasing VMs in the first stage and maximize the expected profit in the second stage. Due to the large number of system states (scenarios) in the environment with multiple services and distributed clouds, the sample average approximation is applied to solve the proposed stochastic programming. Finally, the experiments are carried out based on real workload traces to show the attainable performance of the proposed strategy.