How and Why Netflix Built a Real-Time Distributed Graph: Part 3 — Querying the graph with gRPC…

Netflix details how it engineered the query layer for a real-time distributed graph handling billions of nodes and edges. The system must support two contrasting workloads: wide queries exploring hundreds of connections from one point, and deep queries chaining through multiple sequential hops. Netflix required single-digit millisecond response times on both. The solution combines breadth-first traversal, which processes one level at a time across all nodes, with asynchronous execution to avoid blocking threads on I/O. According to Netflix's TechBlog, this architecture lets the company serve tens of thousands of concurrent queries per second on an actively evolving graph.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton