Re: Big Blue Button

[Ruben Rodriguez]

On 4/27/20 6:38 PM, Jim Garrett wrote:
> To those who were able to attend the BBB test meeting last Friday...how
> did it go?

It was an interesting test, we got to some barely scientific conclusion
that BBB used more CPU than Jitsi with the same number of participants,
but Jitsi was less stable (at least on our instance, which needs
improvement). They have a different feature set so both are worth having.

Some rough estimations of load average, for a number of participants
with webcam and audio:

BBB
Participants - load average
4 - 1
6 - 1.5
8 - 2

Jitsi Meet
Participants - load average
4 - .4
6 - .8
8 - 1.5

A load average of 1 in this case (VM with 8 cores) means that 1/8 of the
total CPU resources were in use. A load of 8 would mean full CPU use.

As a separate datapoint, when using BBB for a class (one presenter with
2 cameras and a screencast, and 20 students with no camera and generally
no audio) got a load of 1.5-2, meaning that the system scales well for
that use case.

Also note that BBB is able to record the session on the server, at no
extra computing penalty. This is the recording from our test:
https://testbbb.fsf.org/playback/presentation/2.0/playback.html?meetingId=183f0bf3a0982a127bdb8161e0c44eb696b3e75c-1587763803652

> Would it be appropriate to create some space on the LibrePlanet wiki to
> record summaries of test results?  It could give people guidance based
> on actual experience.

I think that would be great, please add any info you see fit.