How to Scale Real-Time APIs Beyond 100k Connections Using WebSockets and Redis

Scaling real-time APIs to over 100,000 concurrent connections requires a distributed architecture, as a single Node.js server can handle only 10,000–50,000 WebSocket connections. WebSockets are the preferred choice for bidirectional communication like chat and collaborative editing, while Server-Sent Events (SSE) offer a simpler HTTP-based option for one-way data streams such as live feeds. A key challenge in horizontal scaling is the sticky session problem, where clients connected to different servers may miss events without a shared message bus. Redis Pub/Sub serves as a fast message broker to relay events across all server nodes, though it drops messages if subscribers are slow, making Redis Streams a better fit for critical data. A production-ready architecture typically combines a WebSocket or SSE connection layer, a Redis-backed message broker, and an optional persistence layer for message history.