Building a global and resilient streaming service API
summary
Discovery+ runs a global, multi-region microservice platform that scales from hundreds of thousands to 20 million subscribers and designs for fast recovery not perfect consistency. Services either keep local cached reads or consume Kafka-based event propagation so failures return stale data instead of 500s, and circuit breakers with exponential backoff prevent retry storms. Operational fixes include per-user debug logging and correlation IDs, rate limits to stop the midnight “time bomb,” and capacity tuning like Spring Boot 15s startup and pre-warmed pods.
AI-written summary · about
chapters (4)
related talks
19mHow We’re Scaling Discovery at Netflix Reliably
27mBehind the Scenes: Running Critical Services @ Netflix
19mKeynote: Making and Keeping Netflix Highly Available
18mScaling APIs to Billions: Extreme-Scale Reliability
25mMixing data streams at scale
44mTechniques Netflix Uses to Weather Significant Demand Shifts
01h 08mDystopia as a Service
25mDesigning and Testing Resilient APIs for India Scale