Problems with Audio and Conferences Scalability

Hi,

I have a slightly different situation on my Jitsi Meet installation; at least, I couldn’t find anything like that in the forum.

My server has plenty RAM (64GB) and lots of simultaneous users (sometimes around 500) but distributed in several conferences. I have a 600 Mbps internet link and another with 500Mbps, both full duplex and internet faced (no NAT).

I’m at Brazil and my users are miles away from the server. The quality of the conferences are sometimes awesome, sometimes awful. Lots of problems with audio.

Questions:

  1. From what I’ve read, multiple videobridges would help with my audio issues. Am I right?
  2. If so, would it be possible to use Docker in another server to run all the additional videobridges? The main server and the one(s) for the videobridges would be all connected by a gigabit switch.
  3. I see traffic in port 10000 thru one link and 443 in the other. One link answers for the https request and is the server’s default gateway. I taught that the other link would be idle. How come there are any traffic in this other link? Is it something I can configure?

Hope someone take the time to help me out.

I thank you all for the effort in putting this awesome software available for everybody to use.

I haven’t implemented it yet, but you’d want to look into OCTO. It helps with meetings with people far away from each other.

Thanks @MarcoStine
I’ve read about OCTO but couldn’t understand it quite well. I didn’t realize it could help with the quality of the connections.
I’ll check it out.

Hello @facb69,

I didn’t understand your structure well:

  • Do you reserve one link to UDP/10000 and the other to TCP/443? If so, the audio/video traffic (except the relayed packets) should only be transmitted through a single link. The other link should be mostly idle.

  • Is there a TURN server on the same server?

Hi @emrah
I don’t reserve the IPs.
My URL resolves to IP1 which is the default gateway of the server.
The IP2 should be idle because there’s no DNS pointing to it and there’s no static route or FW rules.

Have you monitored the server about CPU usage in the problem times? Are you using latest stable as there were some UI optimisations for some reports we were seeing about audio issues?

Hi @damencho
My CPU usage at the time are equally distributed thru 32 cores. Load is ok, too.
I was wondering if the great amount of simultaneous conferences on only one videobridge instance would be the cause. I’ll try to add another server with only videobridge pointing to the main one and check the results.
Since port 10000 receives UDP packets it’s not a big issue but I was thinking of a way to avoid unnecessary connections.