Streaming at a distance

[bchah]

Hello friends, what do you think is the best configuration for streaming WebRTC over long distances from a single instance of OME?

I know that in any case the best thing is to have a cluster or instance of OME running closer to the viewers, but I wonder what can be done to increase the resilience of a stream in scenarios of high latency, minor packet loss or other network fluctuations...

In my testing (using Apple Network Link Conditioner) it seems that using UDP with RTX and Ulpfec enabled provided the most resilient playback when the network conditions change. Using TCP relay, the playback would sometimes fall out of sync or start stuttering, and never recover until the player was reloaded. Interestingly in my test scenario it was LL-HLS that fell apart the fastest where I expected WebRTC would have more trouble.

But I am most interested in hearing it from you and the community - what works best for you when streaming to your very farthest viewers?

[getroot]

If the network speed is slower than the bitrate of the stream, there is no way. The only solution is to provide a lower bitrate with ABR.

If the network jitter is high, you need to increase the buffer. This can be helped by increasing the chunk size of Legacy HLS or LL-HLS. As far as I know, there is no effective way to increase the buffer in WebRTC.

If the packet loss is high, TCP connection or WebRTC/Retransmission will be effective.

It seems that your case is a slow network.

[getroot]

For reference, when transmitting video over webrtc in browsers (including libWebRTC), the sender has the ability to adjust the transmission quality (bitrates, resolution, fps) in real time depending on the status of the receiver.

[bchah]

Thank you for your input! I did some testing with the customer and their viewer who was having trouble playing back a Toronto-based stream in Paris. When we tested we were not able to reproduce the scenario of course 😅

The customer is running OME in Docker on Windows - could that environment and the emulation layer be affecting the network conditions perhaps?