We were load-testing Jitsi Videobridge on an Azure VM, steadily increasing the load 10 participants at a time. We peaked at 118 concurrent participants (308 channels, 170 streams) spread out across 12 rooms. The breakup was:
- 10 rooms with 10 participants each, 2 with video, 8 with audio only.
- 2 more rooms with 4 participants each, 2 with vide, 2 with audio only.
After hitting the peak, a large chunk of participants started dropping out and faced issues reconnecting back to JVB.
Our server hosting JVB is a 16-core CPU with 32 GB RAM and two network cards with 30 Gbps bandwidth each and it isn’t maxing out on either of these specs. We spread out our load across more than 20 systems so we believe our systems / network aren’t maxing out either.
We wanted to know if JVB has some limit or breaking point in terms of the number of concurrent participants / channels it can support before it gives away? And if so, what is the best way to scale? Spinning out a new JVB instance?
Could it be that we should be looking at some config / setting to allow more than 100 participants on the server? We found these settings from scouting the forums:
DefaultLimitNOFILE=65000 DefaultLimitNPROC=65000 DefaultTasksMax=65000
Are there any other settings we should be looking at?