Maximum participants with audio only and with both audio video

meet
#1

Hello Jitsi community - I wonder what are the limitation of the webrtc architecture in term of conferencing, I mean in term of participants. If all users are in cellular 3G/4G connectivity, how may people can we have in a call ? If they are on a wifi 10Mbit/sec connectivity?

In indeed the communication is peer to peer and if the speaker has to maintain a connection for each participant, are we limited to 10, 50, 100 or any other number of participants.

Also what would be expected latency when tranporting audio ? video ?

Thanks in advance, Philippe

#2

Peer2Peer is only active when there are two participants, in the moment a third joins is the moment when jitsi-videobridge steps in. Jitsi Videobridge is a SFU and all participants are sending streams to it and the streams are transported where it is needed from there.
The bridge is smart enough to detect link problems for participants and can detect when there is no good bandwidth and will start switching off video streams for it or will send lower resolutions when simulcast is used by the source streams … many smart steps to improve the participant experience.

#3

Thanks for this valuable information @damencho.
If I am using only audio, shall it work in a MESH topology or will still use SFU ?
Also what would be the scalability equation of a SFU unit on on let’s say a AWS t3.xlarge (16Gbyte) server ?
I am asking since the app I need to build will serve university like conferences and may reach 500 participants

#4

So I don’t think mesh is good for more than 3-4 users in a conference.
For our bridges we are using c5.xlarge on meet.jit.si with pretty good load of the service.