Upon some simple test conferences with two participants I had the impression that sound could be delayed more than video. It points out that this is related to cpu load on one of the participants.
Today I managed the first jitsi video conference with own server and 6 participants. Server load (top command) is max. approx. 80% (two cores: full load is represented as 200%)
The bottleneck appeared especially related to the participant’s hardware (and/or connection). Partially the audio was quite bad (slow and/or with intermittent interruptions).
As a consequence, I presented the others my backup audio channel via an own asterisk-based conference (on the same server), via a PSTN gateway.
This worked clearly better (of course with jitsi mic muted). However, some of the participants used their mobile for the phone connection, hence bypassing the internet bottleneck. But… requested audio bandwidth is far less than video. As such it (theoretically) shouldn’t result in timing problems as part of jitsi A/V-stream, assumed that the audio stream can be assigned higher priority over the total connection path.
Any comments/experience notes regarding this? Or even hints of how to get an improved real time experience especially of the audio part of the stream?