Short conversation (Q&A) about causes and mitigation of WebRTC latency. Speakers explain that WebRTC targets ultra-low latency (typically under ~100–200 ms), network path and intermediary servers (TURN, SFU) introduce delay, and remedies include placing servers nearer users and optimizing backbone routing between servers/SFUs.
How do I reduce WebRTC latency?
That's a good question. I mean WebRTC is designed for ultra low latency so typically less than 100200 milliseconds and where latency comes in is typically the network and that means typically any server in the middle. So if you for example have a call that's related via your turn server, then the time it takes to travel from one user to the turn server to the other user is what defines your latency.
Okay. And reducing it means placing machines closer to the users and then enforcing on understanding and optimizing the routes within the backbone of your infrastructure.
Yes, which is why we see people who run turn servers actually they run large turnover networks. If you were talking about sfu, you typically connect user to the closest sfu, then have that SFU connect to another SFU and then to the other user.
Okay.
The connection between those two is typically on the backbone of cloud provider so it's more reliable than general Internet. It's all old stuff, 2015 or so.
Okay.
Meta rebuilt AV1 for mobile scale. Should smaller teams follow? | Chats with Arin
