Infrastructure efficiency: How a classic math problem helped reduced our capacity buffers

This post details the application of the newsvendor problem and statistical modeling to optimize capacity buffers in a private cloud infrastructure. It introduces the critical fractile formula (CF = Cu / (Cu + Co)) and its application to infrastructure by defining underage cost (SLA breaches, emergency procurement) and overage cost (stranded hardware). The post elaborates on key factors for successful implementation: establishing high-fidelity demand signals, strategic aggregation, quantifying organic variability, leveraging resource fungibility, navigating lead-time volatility, and anchoring to service SLAs. It specifically highlights the challenges and application within LinkedIn's stream-processing platform, contrasting its needs with batch processing.