Online Rack Placement in Large-Scale Data Centers via OSO

Saumil Baxi et al. present an online sampling optimization (OSO) algorithm for online rack placement and report a software deployment across Microsoft’s data centers.

• Main announcement: The paper introduces OSO (Online Sampling Optimization) and an online integer optimization model for rack placement to maximize demand coverage under space, power, and cooling constraints; the authors report that the algorithm was packaged into a software solution deployed across Microsoft’s data centers and integrated into a human-machine interactive decision process.
• Background & evidence: The paper (submitted 22 Jan 2025; last revised 16 Jan 2026) includes theoretical guarantees (strong competitive ratio, sublinear regret) and provides empirical deployment analysis where econometric tests estimate a 1-3 percentage point reduction in power stranding; full paper and artifacts available via arXiv PDF, TeX source, and DOI.