Glossary
SFU (Selective Forwarding Unit)
An SFU, or Selective Forwarding Unit, is a media server that receives WebRTC streams from each participant in a group call and forwards the appropriate streams to each other participant without decoding or re-encoding the media, enabling scalable multi-party video calling.
General definition
In a group WebRTC call, peer-to-peer connections between every participant become impractical beyond two or three people. An SFU sits in the middle: each participant sends one upstream and receives individual streams from the SFU for each other participant. The SFU selects which stream or quality layer to forward per receiver without performing the expensive step of decoding and re-encoding.
- Low server CPU: no re-encoding means the SFU processes only packet routing
- Per-subscriber quality selection: the SFU can forward different simulcast layers to different receivers
- Participant-level control: individual streams can be muted, scaled, or dropped independently
- Lower latency than MCU (Multipoint Control Unit) approaches because there is no transcoding delay
Popular open-source SFU implementations include LiveKit, mediasoup, Janus, and Jitsi Videobridge. An SFU is the standard architecture for modern group video calling platforms. It differs from a MCU, which mixes all streams into a single composite video, at the cost of much higher server CPU.
In the Ethora ecosystem
Ethora’s voice and video calling is built on WebRTC with an SFU architecture, enabling group calls with per-participant stream control and simulcast adaptive quality. Each participant sends one upstream and receives individual streams, keeping call quality high for all participants regardless of the group size.
Teams deploying Ethora on their own infrastructure manage the SFU alongside the rest of the communication stack, giving full control over media routing policies, recording pipelines, and the ability to run the SFU in the same network region as the chat and AI services for minimum jitter and latency.
