below are my setup,
a) jvb2 + octo
b) 1 x jicofo on AWS EC2 c5n.9xlarge (vCPU: 36, Memory: 96GB, network: 50Gbps)
b) 5 x bridge server on AWS EC2 c5n.9xlarge (vCPU: 36, Memory: 96GB, network: 50Gbps)
c) simulcast: ON, ChannelLastN: 10, disableAudioLevels: True, DISABLE_VIDEO_BACKGROUND: true, enableLayerSuspension: true
The octo is working perfectly and distributed participants evenly across 5 bridge servers. however, when more than 230 participants come into single conference then it become unstable. some participants will be kicked out, and some can’t hear/see screen sharing. based on my observation the bottleneck is not at server backend, but the front-end browser cannot cope with the streams forwarding.
seem like the breakpoint is between 230 ~ 240 participants / conference. 1 more observation, when there is more than 4 participants with video ON or screen sharing then CPU fan will spin like crazy.