Media servers need to account for client-side CPU in WebRTC

27 Jan 2026 · 2 min
1:55

In this presentation, the speaker argues that while SFUs (Selective Forwarding Units) already manage network conditions for WebRTC calls, they must also orchestrate client‑side CPU load. To do this, clients should continuously send CPU telemetry to the server. The SFU can then make intelligent, per‑participant decisions such as switching to lower‑bitrate or different simulcast/SVC layers, applying temporal scalability to reduce frame rate, stopping or rerouting video streams based on layout and context, and renegotiating bitrates. These server‑side actions aim to improve overall media quality by tailoring resource usage to each participant’s device capabilities. The talk concludes by asking the audience to identify the single most valuable piece of client‑side CPU telemetry for the server to act upon.

Chapters

More from Tsahi

9:15
youtube.com

After 15 years of WebRTC, here is when I drop peer-to-peer

1mo ago
1:39
youtube.com

Why UDP won the real-time war #programming #webrtc #shorts

1mo ago
1:37
youtube.com

The Difference Between SFU And TURN Servers #WebRTC #RealTimeCommunications

2mo ago
1:43
youtube.com

Why server logs are useless for WebRTC #webrtc #getStats #observability

2mo ago
1:49
youtube.com

How does WebRTC even know a packet was lost?

2mo ago
1:42
youtube.com

WebRTC gives you every tool to watch your calls, then collects nothing

2mo ago
1:40
youtube.com

All green but calls are broken #qoe #dashboards #webrtc

2mo ago
14:03
youtube.com

Meta rebuilt AV1 for mobile scale. Should smaller teams follow? | Chats with Arin

2mo ago
1:16
youtube.com

Is your video API vendor ready for AI agents?

3mo ago
1:21
youtube.com

What "$4 per 1000 Minutes" Really Costs You (Video API Pricing)

3mo ago
1:30
youtube.com

How to Read the WebRTC Tools Hub in 90 Seconds

3mo ago
See all content