Efficient Cloud Server Deployment Under Demand Uncertainty.

This article focuses on the cloud server deployment problem faced by large cloud providers such as Microsoft Azure, addressing the challenge of planning hardware deployments under uncertain future demand while minimizing costs. The authors formulate this problem as a two-stage stochastic mixed integer optimization program that jointly considers long lead times for infrastructure preparation ("row building"), supplier selection, shipping, and deployment scheduling, subject to temporal, capacity, and throughput constraints. They identify two key structural properties—hierarchical throughput constraints and demand compatibility homogeneity—that enable a tight convex relaxation of the second-stage problem, allowing efficient solution via a hybrid Level-Benders decomposition algorithm. Using real production data from Microsoft Azure, the proposed algorithm demonstrates significant cost reductions compared to deterministic and scenario-based baselines, maintains robustness across varying demand uncertainties and supply availabilities, and accommodates different risk preferences including expectation, conditional value-at-risk (CVaR), and mean-deviation measures. The study highlights the benefits of explicitly modeling demand stochasticity and risk aversion in cloud supply chain operations and provides a foundation for further extensions in cloud infrastructure management.