Why GraphQL gets N+1

Bea asks for posts { title, author { name } } — one HTTP request for 50 posts. Sven watches the database log light up: 51 queries. The schema is correct. No bug was introduced. This is GraphQL working exactly as designed, and it is costing 600 ms instead of 80 ms.

How GraphQL executes a query

GraphQL resolves a document by walking it depth-first and breadth-first at each level. For every field the client selected, the server calls the resolver registered for that Type.field pair. Resolvers receive (parent, args, context, info) and return a value or a Promise. They are deliberately isolated: each resolver does not see its siblings, does not know how the parent fetched its data, and cannot inspect what other resolvers will run next.

This isolation is what makes GraphQL composable — you can add a field to a type without touching any other resolver. But it is also the direct cause of N+1.

Why the isolation causes N+1

Consider the query posts { title, author { name } }. The execution engine:

1. Calls Query.posts → runs SELECT * FROM posts LIMIT 50 → returns 50 post objects.
2. For each of the 50 posts, calls Post.author(parent=post, ...) → runs SELECT * FROM users WHERE id = post.author_id → 50 separate queries.

Total: 51 database round-trips for one HTTP request.

A REST controller writing the same endpoint would naturally write SELECT posts JOIN users ON users.id = posts.author_id — one query. The developer sees the whole response shape in one place and can optimize the SQL. In GraphQL, the Post.author resolver sees only one post at a time. There is no global planning point.

The waiter metaphor

A restaurant where each table orders items one at a time. A table of six sends the waiter back six times — once per item. The kitchen is fast, but the waiter spends the shift walking. The kitchen’s speed does not matter when the walk is the cost.

DataLoader is the manager who says: hold tickets until I clap, then bring them all at once. The clap is one event-loop tick. This lesson explains the problem; the next two explain the fix.