Service-to-service GraphQL: The New Sweet Spot!
summary
Argues that GraphQL works well for service-to-service when you have multiple external and internal clients, federation with resolvers running inside services, and workloads like SDUI or LLM-driven flows that need auth-preserving, field-level fetching. Recommends a single GraphQL schema with auth directives or permission logic to avoid duplicate REST and GraphQL implementations. Shows numbers: 20KB JSON → 2.8KB with ZSTD, Protobuf encodes ~15KB then compresses to ~2.5KB for only a 300 byte advantage for Protobuf when compressed, deserialization times are sub-millisecond (Protobuf in Go fastest). Mentions Simdjson, ZSTD, Argo as practical tools.
AI-written summary · about
chapters (10)
- 0:00Introduction and context
- 1:23Why GraphQL for services
- 3:50Use cases for service clients
- 6:59Avoiding duplicate APIs
- 9:49Single GraphQL definition benefits
- 11:35JSON performance myths
- 14:45JSON parsing optimization
- 17:56Performance trade-offs
- 19:00Binary formats for GraphQL
- 20:05Conclusion and summary
related talks
30mWhy GraphQL vs. Why not GraphQL
51mWhy GraphQL is Perfect for Microservices
27mWhy GraphQL is Perfect For Microservices
23mHarnessing the Power of GraphQL
37mGraphQL and Microservices or How we learned to be big and small at the same time
48mGraphQL Is the New Sexy
27mA no nonsense GraphQL and REST comparison
15mThe Right Size For GraphQL