How we deliver 1Tbps of video with our WebRTC peer-to-peer network
summary
Streamroot’s WebRTC DNA delivers 1.2 Tbps peak of peer-to-peer video and offloads about 70% of CDN traffic by matching viewers via a router/tracker system, handling >600k concurrent users and huge join spikes. Implementation uses WebRTC data channels, Go backend, Redis counters, nginx load‑balancing on Google Cloud Kubernetes (daemonsets, tuned conntrack), a router service to shard trackers, chunked segments (60KB chunks), and compressed signaling yielding ~10x reduction; the system scales and met production World Cup demand.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:46About Streamroot
- 1:45Advantages of Streamroot
- 2:48How does Streamroot work
- 8:45Peertopeer sharing
- 11:02Other optimizations
- 14:29Football World Cup
- 15:48Results
- 17:13Scaling Journey
- 18:09Architecture Changes
- 21:02Customer Dashboard
- 22:53Capacity Graph
- 24:45Network Path
- 26:36Optimizations
- 27:51Connection tracking
- 28:59Problems
- 40:32User optin
related talks
20mBeyond P2P: Video Routing in WebRTC
17mLive Video Streaming as "tail -f"
14mMovienight 101: How to Power a Video Streaming with Kubernetes and WebRTC
22mWebRTC: Connecting Embedded, Mobile and Desktop with the Lowest Latencies
43mThe unstoppable secret superpower of WebRTC
27mWebRTC does what? Learn how robotics subsecond broadcast and game streaming is powered
32mWebRTC Update
20mDecentralizing Video Streaming