realtime-systems

Choosing between WebSocket and SSE for real-time communication depends on whether you need bidirectional messaging or simple server push. WebSocket supports full-duplex interactive sessions, while SSE is ideal for unidirectional live updates with lower overhead.

Scaling real-time presence across multiple instances requires a Redis-backed pub/sub architecture to broadcast state changes and a presence TTL mechanism to automatically expire stale user connections without overloading the network.

CRDT-based collaboration handles live synchronization in distributed apps by allowing concurrent edits without strict central locking. Conflict-free replicated data types merge changes deterministically, ensuring all nodes eventually reach consistent state.

You need Redis or NATS for horizontal scaling of real-time WebSocket systems to route cross-instance messages. Without a pub/sub message broker, separate server instances cannot share state or broadcast events to clients connected to other nodes.

Common anti-patterns when building real-time systems with pub/sub include missing heartbeats that cause false connection drops, ignoring backpressure leading to memory bloat, and lacking presence TTL which results in stale user state synchronization.

SSE provides built-in browser reconnection for fault tolerance in live syncing, but managing backpressure requires server-side handling. You must throttle or drop events when slow clients cannot process the real-time stream fast enough.