Cost Optimization in Multi-site Multi-cloud Environments with Multiple Pricing Schemes

Abstract

The rapid adoption of cloud computing has led to the proliferation of cloud providers, along with increasing options offered by them. Currently cloud providers offer a wide range of services (machine sizes, availability modes, storage etc.) with complex pricing schemes (spot pricing, reservation pricing, etc.). At the same time, customers require distributed deployments in order to meet their own SLA commitments. This has led to the concept of multi-site multi-cloud deployment schemes that allows enterprises to deploy highly performant distributed applications such as gaming using multiple cloud providers. Existing research has solved the problem of cost optimization of such deployments BUT under the assumption of a single pricing scheme for all providers. The introduction of varied and dynamic pricing introduces several complexities that needs a fresh look at this problem. In this paper we present a solution for mapping the compute and storage requirements of a set of related enterprise sites onto cloud providers who offer multiple, dynamic pricing schemes. Our goal is to optimize the cost of a multi-site deployment consisting of compute and storage components at each site while meeting SLA requirements of the application. We show that our approach can achieve cost reductions of up to 22% for the customer even for medium sized deployments consisting of tens of sites over the scheme that uses either a single cloud provider or fixed pricing schemes.