Infrastructure Week 2023 -- Simulcast

Описание: Read the technical deep-dive on Simulcast:
https://www.daily.co/blog/simulcast/

Or, explore the previous videos in our infrastructure series:
(1) Infrastructure series overview:    • Welcome to Infrastructure Week 2023 at Daily  
(2) Global Mesh Network:    • Infrastructure Week 2023 -- Global Mesh Ne...  
(3) Interactive Live Streaming:    • Infrastructure Week 2023 -- Interactive Li...  

First off, what is simulcast and why is it important?

There are a few different designs for WebRTC call servers. One of these is simulcast. Simulcast is WebRTC’s secret weapon. It’s a way to have a bunch of people on a call together and ensure that they’re all having the best experience possible. It makes sure that call participants are seeing the best video quality their internet connection allows, but not using more bandwidth or processing power than they may need to.

How does simulcast work?

Simulcast allows individual participants to create different quality versions of their own video to send to the server. In doing so, the server forwards only the right quality of video to each person. This means that even if someone on the call is experiencing poor network connectivity, only that individual’s video will be affected, not that of everyone else in the call.

In other words: simulcast retains the overall call quality, compared to the alternative that we’ve all experienced: one person has bad connectivity, resulting in poor call quality for everyone else.

A sneak peek underneath the hood of simulcast

Video layers are at the heart of simulcast. Layer 0 is the lowest quality, and it goes up to layer 2 for the highest. And something chooses which layer each person receives for each video track. That choice can be made automatically or manually, and it can be made, to an extent, by the SFU or the participant.

The WebRTC standards help us out with the automatic part. There’s a mechanism built into the protocol itself to monitor network throughput over each of these connections. The server can use that data to change which layer it’s choosing to send to a particular participant. So Bob can get the right video without you having to do anything. But, you can also control this from the participant end.

You can learn more about about this and more in our comprehensive blog post on simulcast:
https://www.daily.co/blog/simulcast/