Explain:

• Real-time audio/video

• Low latency (<200ms)

• Packet loss, NAT, firewalls

• Scaling from 2 → 10,000 users

Key insight:

Video calls are not just “streaming” — they’re distributed real-time systems.

What Is WebRTC? (Foundation Layer)

WebRTC = real-time communication protocol

Handles:

• Media capture (camera/mic)

• Encoding (VP8, H.264, Opus)

• NAT traversal (STUN / TURN)

• Secure transport (DTLS, SRTP)

⚠️ Important:
WebRTC does NOT define architecture (P2P / SFU / MCU).
It only provides the pipes.

Architecture #1 — Peer-to-Peer (Mesh)

How it works

Each participant sends streams to every other participant.

Pros

• Lowest latency

• No media server cost

• Simple to implement

Cons

• Bandwidth explodes (O(n²))

• CPU overload on clients

• Breaks beyond 3–4 users

Real-world use

• 1-to-1 calls (WhatsApp, FaceTime)

• Small internal tools

❌ Not suitable for meetings

Architecture #2 — MCU (Multipoint Control Unit)

How it works

• All streams go to a central server

• Server decodes, mixes, re-encodes

• Sends one combined stream back

  

Pros

• Very low client bandwidth

• Simple clients (mobile friendly)

• Easy recording

Cons

• Extremely CPU expensive

• High server cost

• Added latency

• Poor scalability

Real-world use

• Legacy enterprise systems

• Hardware conference rooms

• PSTN bridges

❌ Rarely used today for large-scale apps

Architecture #3 — SFU (Selective Forwarding Unit) ⭐

How it works

• Clients send streams to SFU

• SFU does not decode

• Forwards streams selectively

  

Clients decide:

• Which streams to render

• Which resolution to request

Pros

• Scales to thousands

• Low server CPU

• Flexible layouts

• Adaptive quality

Cons

• Higher client complexity

• More client bandwidth than MCU

Real-world use

✅ Zoom
✅ Google Meet
✅ Discord
✅ Microsoft Teams
✅ Webinars & live classrooms

⭐ SFU is the industry standard

Why SFU Wins in the Real World

Real-World System Design (Zoom / Meet-Style)

Components:

• Signaling Server (WebSocket)

• SFU Media Server

• TURN servers (fallback)

• Load balancer

• Region-aware routing

Flow:

1. User joins meeting

2. Signaling server exchanges SDP

3. Media flows to SFU

4. SFU forwards streams

5. Active speaker detection

6. Dynamic bitrate adaptation

References & Further Reading

📚 Core WebRTC & Architecture References

1. WebRTC Official Documentation
   Best for fundamentals, APIs, codecs, and transport details
   → Covers ICE, STUN/TURN, DTLS, SRTP, media pipelines

2. IETF RFCs (WebRTC Standards)

   • RFC 8825 – WebRTC Overview

   • RFC 8834 – Media Transport and Security
     Best for protocol-level accuracy

🏗️ SFU / MCU Conceptual Deep Dives

3. Jitsi Architecture Docs
   Excellent real-world SFU explanation (Jitsi Videobridge)
   → Explains selective forwarding, bandwidth adaptation, scaling

4. Janus WebRTC Server Documentation
   Great for understanding SFU internals and plugin-based design

5. mediasoup Design Docs
   Modern SFU used in production systems
   → Clear explanations of producers, consumers, routers

🎥 Real-World Product Engineering Blogs

8. Zoom Engineering Blog
   Explains why SFU-like architectures scale better than MCU

9. Google Meet / WebRTC Engineering Talks
   Covers simulcast, congestion control, and adaptive streaming