Real-Time Communication Platform

Problem

Operators and end users needed live updates across sessions — presence, notifications, and state sync without polling-heavy backends.

Architecture approach

WebSocket gateway with Redis pub/sub backplane; domain events propagate state changes to connected clients and async processors.

Key challenges

• Connection scaling and sticky session trade-offs
• Ordering guarantees vs delivery speed
• Graceful degradation when realtime path fails

Scalability decisions

• Horizontal gateway nodes with shared pub/sub
• Client-side reconnect with cursor-based catch-up
• Fallback to polling for critical read paths only

Outcome

Operators see live state with acceptable latency under load — architecture documented for regional expansion.

Tech stack