Lessons from Scaling GraphQL to Half a Billion Requests per Minute
summary
GraphQL powers a platform that grows from thousands of RPS to half a billion requests per minute and the team shrinks GraphQL EC2 from ~50,000 cores to ~6,400 cores by micro‑optimizing Node.js, right‑sizing instance types, aggressive benchmarking at 4x expected load, and then replacing handwritten resolvers with a Rust runtime called Tailcall that adds an ahead‑of‑time analyzer and JIT planner. Federation would have added ~30% more compute; the config‑driven runtime preserves performance, reliability, and maintainability and is production‑proven and open source.
AI-written summary · about
chapters (7)
- 0:00Scaling GraphQL at Dream11
- 4:30Performance challenges
- 10:05Reliability and constraints
- 14:54Maintainability struggles
- 16:45Refining GraphQL architecture
- 20:45Introducing Tailcall
- 26:45Key takeaways and conclusion
related talks
20mScaling GraphQL on AWS: Production Architecture for High-Volume Data Systems
39mGraphQL Accelerated
41m'Divide et impera' with GraphQL and SSR
46mMoving to GraphQL
23mHow GraphQL Helped Create Scalability and Stability in the Retirement Space.
23mHow to edge cache GraphQL APIs
26mBuild a performance-optimized GraphQL API
25mOptimizing GraphQL Applications