Live streaming differs from video-on-demand:
Latency targets:
- Standard live: - seconds (HLS/DASH)
- Low latency: - seconds (LL-HLS, WebRTC)
- Ultra-low: < second (WebRTC only)
Architecture:
- Ingest: Streamer sends RTMP to edge server
- Transcode: Real-time encoding to multiple qualities
- Package: Segment into - second chunks
- Distribute: Push to CDN edge nodes
Challenges: No buffering ahead. Must process faster than real-time. Handle viewer spikes instantly.
Chat integration: Separate WebSocket service. Eventually consistent with video.