Jitsi Videobridge: the state of the art SFU that powers Jitsi Meet
summary
Jitsi Videobridge is a decade‑old SFU (selective forwarding) that runs as open‑source servers and Docker containers and can be hosted via JaaS; it forwards packets (no transcoding), supports simulcast and SVC, chooses layers by client constraints plus an SFU bandwidth estimate, fills tiles then upgrades by speaker order, and uses RTP probing/RTX duplicates to ramp bandwidth. Cascading SFUs reduce latency and balance load; a multi‑tier mesh targets 10,000 receive-only endpoints. Implementation details: Java/Kotlin with a memory pool in Java heap to keep GC pauses low and a public limit around 500 participants.
AI-written summary · about
chapters (9)
- 0:00Introduction to Jitsi Videobridge
- 2:58History and evolution of JVB
- 4:46SFU basics and codec support
- 9:03Bandwidth allocation strategies
- 14:02Cascading SFUs explained
- 17:44Scaling to 10,000 participants
- 20:38Q&A session part 1
- 26:02Q&A session part 2
- 29:36Q&A session part 3
related talks
20mBeyond P2P: Video Routing in WebRTC
15meXpressSFU: Toward Super-Scalable Video Conferencing with SmartNICs
19mScaling for Large Group Calls
28mBandwidth Estimation in the Janus WebRTC Server
17mScaling Decentralized Low Latency Infrastructure with Huddle01
44mSeamless mobile real-time communication with WebRTC
28mPipelining WebRTC videoconferences in Elixir with Membrane Framework
41mServers for WebRTC: It is not all Peer to Peer