JavaEdge

Dec 8, 2024 · Backend Development

Netflix’s Service‑Level Priority Load Shedding: Protecting User‑Initiated Requests

This article explains how Netflix extended its priority load‑shedding strategy from the API gateway to individual services, detailing the classification of user‑initiated versus pre‑fetch requests, the implementation of partitioned concurrency limiters, CPU‑ and I/O‑based shedding, test results, and real‑world impact on availability.