Downtime Distribution, Risk Volatility, and Optimal Backup VMs Allocation in Cloud Services

Service Level Agreement (SLA) is an important basis to ensure the availability of cloud service. Cloud service providers seek to not only fulfill the SLA, but also simultaneously minimize the risk uncertainty. Under such a mechanism, service providers look forward to determine the optimal number of backup virtual machines (VMs) to minimize operational costs. In this paper, we propose a data-driven risk measurement method under multi-stage SLA and empirically portray the likelihood and risk values of compensation events at different stages under several service periods. Furthermore, a multi-objective optimization function is established to balance the provisioning cost and risk volatility for cloud service providers with different risk preferences. This paper provides a practical method to cloud service providers by addressing the risk uncertainty in multi-stage SLA from cloud computing resource management perspective.