GraphQL at Facebook
summary
Facebook claims 300 billion queries a day and a schema of nearly 10,000 types. The stack centers a single source of truth: objects expose a canonical gen(viewer,id) that enforces authorization by returning null for unreadable items, enables batching via DataLoader to avoid N+1 round trips and provides server-side caching, and emits opaque global IDs (base64 of type:localid) that let clients cache and refetch without learning backend details.
AI-written summary · about
chapters (16)
- 0:00Start
- 3:28GraphQL: How?
- 4:07How do I make GraphQL efficient?
- 4:38Facebook's GraphQL Stack
- 5:48Think Graphs, not Endpoints
- 8:43Business Logic
- 9:13What is a Todo Item?
- 10:48Implementing Authorization
- 14:43Authorization in GraphQL
- 17:19Efficiency
- 18:41Implementing Batching
- 19:03DataLoader
- 20:42Caching
- 22:46What do we want in IDs?
- 23:07Refetch Identifier
- 25:48Single Source of Truth
related talks
27mData fetching for React applications at Facebook
47mIntroduction to Graphql
40mGraphQL: Designing a Data Language
48mGraphQL - a type system for your API
11mA Brief History of GraphQL
01h 00mGraphQL - A query language to empower your API consumers
30mBest Practices for GraphQL and GraphQL Subscriptions at Scale
27mExploring GraphQL