Stochastic programming models for provisioning cloud computing resources

Official URL: http://risc01.sabanciuniv.edu/record=b1664844 (Table of Contents)

In this study, we focus on the resource provisioning problem of a cloud consumer from an Infrastructure as a Service (IaaS) type of cloud which could be deployed as on-demand or could be reserved in advance. Even though the hourly usage cost of the reserved instances is smaller than that of the on-demand instances, the inherent uncertainty in demand and price makes it attractive to complement a base reserved capacity with on-demand capacity to hedge against spikes in demand. We first formulate the cloud resource provisioning problem as a risk-neutral dynamic multistage stochastic program, which serves as the base model for further modeling variants. In this model, decisions are made dynamically over time by taking into account both the realized uncertainty and previous decisions at a given decision epoch. To accentuate the value of dynamic modeling, we transform the base model into a static one by deciding on all reservation amounts at the start of the planning horizon without observing the realized uncertainty. Finally, chance constraints integrated into the base formulation require a minimum service level met from reserved capacity, provide more visibility into the future available capacity, and smooth out expensive on-demand usage by hedging against possible demand fluctuations. Two alternate modeling paradigms – node-based versus scenario-based – are applied to all formulation types, and the corresponding computational efficiency is explored in experiments. Furthermore, the solution structure is also investigated in our numerical study with the goal of providing managerial insights.