Netflix built a Real-Time Distributed Graph (RDG) to unify member journeys across streaming, gaming, and ads. Traditional microservices siloed data, making cross-domain analysis difficult. Graphs were chosen for fast relationship traversal and flexibility as new connections emerged, replacing slow, manual data stitching.

The system ingests events via Apache Kafka, processing up to a million messages per second per topic. Apache Flink jobs filter, enrich, and transform these into graph nodes and edges, publishing to Data Mesh for storage. Netflix moved from one monolithic job to a per-topic model for better tuning and scalability.

Traditional graph databases like Neo4j couldn't scale horizontally for Netflix's real-time workload. The RDG uses a property graph model, storing billions of nodes and edges. This architecture enables low-latency queries for personalized experiences, a critical need as Netflix expands into live events and gaming, requiring a unified context engine.