We are trying to broadcast from HD4 with 4 active sim using webRTC. But observed that the actual video thruput is only 1/3 of a single carrier traffic. What’s the recommended config for it?
4 sims are of different carrier and similar bandwidth, and were set to the same priority.
WAN smoothing: seems not needed for one way streaming, tried Normal, didn’t see improvement.
FEC: seems good for one way streaming, we tried LOW or HIGH, didn’t see improvement.
Receiving Buffer: seems useful when packet loss / out of order is happening? Tried, didn’t see improvement.
Latency Diff Cutoff: used to prevent slow links from slowing down the overall traffic. Tried, didn’t see improvement.
Expected behavior:
Should see bandwidth bonding, higher bitrate & more reliable streaming from 4 sims.
Observed:
Data are distributed on 4 carriers. 100 ~ 200 kbps each.
Only 500 kbps total bitrates from webrtc. Inconsistent bitrate, low video quality.
I’m wondering if webrtc is not working well with multiple path, or there is any misconfig we had.
WebRTC based video should work fine, we have many customers running vmix call sessions over PepVPN / Speedfusion without issue at reasonably hight bitrates.
All those things you turned on and off one at time can often work well together, WAN smoothing in particular though can make a huge difference if you have unreliable connectivity and the bandwidth to support it.
Some things to consider -
Where is the hub hosted?
What kind of connectivity is available at the hub?
What does the connectivity look like if you consider the routing from the hub to each cellular WAN in terms of quality of those paths? - any evidence of loss / high jitter / high latency between the cellular links and the hub?
What does the connectivity look like if you consider the traffic path from the hub to whatever video platform you are using if this is being ultimately piped out to some CDN or something like that is that traffic path looking clean?
You say all of the cellular WANs are of similar performance, what is that kind of performance? If you use the PepVPN test tools what sort of throughput do you see across the tunnel? I would test using both that and the WAN analysis tool testing to and from the hub which lets you prove that the individual links can carry sustained traffic both inside and outside of the PepVPN tunnel.
You may also want to test what happens with various combinations of your 4 cellular WANs - carriers can and do sometimes share infrastructure / backhaul from a mast, so there is potential you may just be contending with yourself. Sometimes less is more
500Kbps bitrate is pretty low, is that all you are expecting the video to send or should that be much higher?
Multi-carrier has more packet loss even when the network condition is good, not sure if it’s because the antennas are having signal interfering to each other because they are sending packages at the same time. Even if we turn on FEC or Wan Smoothing, it just has more packet loss than single cell use case.
Our WebRTC client’s ABR (adaptive bitrate) logic is pretty sensitive to packet loss.
As a result we tune the ABR logic to be less sensitive to packet loss and improved the video streaming experience.
I’m actually experiencing the same exact thing with my Max BR2 Pro 5G.
I cant get a reliable 10mbps stream out without massive issues every couple minutes.
4 cellular connections, 2 internal on T-Mobile, 2 external ATT nighthawk m6 pro.
When testing each individual connection on SFC i get 60-100 down and 25-50 upload.
When i combine them all together, and test through the SFC TCP speed test, i get 36 down, and 30 up. Not sure why my speeds are so much lower on SFC than just one individual connection. Is it not bonding the bandwidth? It seems that instead of adding more bandwidth with additional connections, i am actually getting less throughput to SFC.
I’m using the SFC-Atlanta location, and I am in Atlanta. The latency on all 4 connections generally stays between 4-80, sometimes spiking to low 100s. I have my own streaming server deployed in Atlanta 26 miles from the SFC data center. ping time is around 4ms. I did traceroutes and all look pretty clean, as clean as AT&T DNS lets it be, haha.
I know that I’m connecting to two different sets of towers with the two carriers, and both have significant backhauls. I have USB stick LTE modems that are getting 100/35 on any carrier here.
At this point it seems that really the issue happens with the bonding in SFC. like he overhead is way too much and the bonding isn’t pushing out the stream bits persistently. About every 1-3 seconds, the throughput speeds just drop to 0 and then come back to normal a second later.
