How would you design a WebSocket-based chat system at scale?

A scalable chat system uses WebSockets for real-time messaging, a message broker for distribution, and persistent storage for history.

At scale, a chat system must handle millions of concurrent connections. WebSocket servers manage live connections, while a pub/sub system like Redis or Kafka ensures message distribution across nodes. Messages are persisted in a database for history. Load balancers with sticky sessions ensure users remain connected to the same node, while fallback reconnection logic ensures resilience.