How would you design a Zoom/Google Meet signaling system using WebSockets?

A signaling system uses WebSockets to exchange session metadata (SDP, ICE candidates) before peer-to-peer media connection is established.

In video conferencing systems, WebSockets are used for signaling only, not media transfer. Clients connect to a signaling server, authenticate, and exchange SDP offers/answers and ICE candidates. The system must support room-based routing, low latency message delivery, and horizontal scaling via pub/sub (Redis/Kafka). Once signaling completes, media flows directly via WebRTC peer-to-peer or SFU (Selective Forwarding Unit).