Jitsi for large number of users

Hi,

Is it possible to implement Jitsi videobridge for about 1000+ concurrent communication, each session having 20-30 participants. The parent application is a PWA developed using ReactJs / Node.js.
Also what would be the approximate hardware requirements of the server, say for 10 or 100 conferences each with 20-30 participants. The ideal resolution is 720px, but at times would require 1080p.

Thanks in advance,
Neelakantan G

How much bandwidth does everyone have? (short answer: Nope.) You probably want to use Zoom.

Zoom receives video streams from each user and then combines them into one video stream that gets sent back to each recipient. This cuts down on bandwidth but increases server processing. It also means that the room size is limited to the server’s capacity; more people doesn’t increase the bandwidth requirements for any recipient.

Jitsi is strictly a video relay system. If there are only two people, then you talk directly (no relaying needed). This means that 1-1 communications are encrypted and even the server can’t intercept it.

But if there are 3+ attendees, then the video streams are relayed through the Jitsi server to everyone. This means that there are very few resources necessary on the server (CPU and RAM), but there is more bandwidth to each recipient and the amount of data increases with each conference attendee. If one person transmits 2Mbps of video, then 8 people consumes 16Mbps. This may not sound like much if you have a gigabit connection, but if you’re on a smartphone in the United States, then 6 people will probably max out your bandwidth. (Don’t use Jitsi at home if someone else is streaming Netflix or some other video service. You won’t have the bandwidth for both.)

There is also the impact to the client’s resources. 1-1 video requires the client to decode and display one video stream. A room with 8 other people means your client must decode 8 concurrent video streams. More than 6 people might be too much for an older smartphone. (Unlike Zoom, where the client only receives one video stream that contains everyone already it.)

6 Likes

This has got to be the shortest yet most complete comparison of Zoom vs. Jitsi resources use.

@damencho @saghul would you say this is accurate ? Any further comments on capacity planning ?

2 Likes

Just compared the resource demand of a Zoom Meeting (8-11 participants) to a meeting on Jitsi (4-6 participants) with the help of my system monitor (Linux Mint, i5-2400 3,1 GHz Quadcore).

My results suggest, that relative CPU impact of Jitsi exceeded impact of Zoom by factor 5, while bandwidth demand might even have not significantly differed at all.

I calculated relative CPU usage per 100 kiB/s Download. Please note, that bandwidth values cannot be directly compared, since I did not take participants (variable) video resolutions into account.

Interestingly absolute CPU usage in Jitsi meeting never fell below 17 %, even if all 6 participants turned audio and video off.

[Jitsi meeting was on meet.adminforge.de instance. All participants used the Jitsi Meet App. We also used the App in the Zoom Meeting.]

1 Like

@michael-j - These findings are interesting. Thanks for sharing. But I wonder how fair the comparison is if the baseline is not the same. It’d be interesting to see what the bandwidth impact is with the same number of clients/participants (5 additional participants is perhaps too significant to ignore).

Also, per Jitsi CPU utilization, did you consider the known performance improvements (disable audio meter, focus indicator e.t.c…)? Those have been known to have very significant impacts on the CPU utilization. I wonder how that would impact the comparison between both platforms.