Jitsi architecture

Hi all,

We are aiming to deploy jitsi for an entity with more than 100k users, all feedback on those kind of use case are welcome.

We have installed 3 hosts on [CentOS](file:///home/user/Documents/Notes/CentOS.txt) 7 as follow (this is a testing architecture) :

  • One meet jitsi on the frontend with jicofo, prosody and nginx
  • 2 separate Videobridges servers with [PubSub](file:///home/user/Documents/Notes/PubSub.txt)

We have some questions about benchmarking.

  • We would like to know with a gigabit connection, how many users can connect on a single room ?
  • Which is the best tool to benchmark it automaticaly.

For example on a 25 MB connexion we can be 4 users with audio and video, is it normal ?

Thank you

There are number of posts about bandwidth in the forum, you can check for example this and estimate it yourself cause there are many variables in the equation:

Hello,

Thank you.

What the best tool to benchmark it or to strees it ? there is jitsi hammer on github but the last commit was a long time ago.

We are currently using selenium and chrome to test the new bridge, for now this is the way to go.
We are using this test https://github.com/jitsi/jitsi-meet-torture/blob/master/src/test/java/org/jitsi/meet/test/MalleusJitsificus.java that can create X number of conferences with Y number of participants using a selenium grid.

Hi damencho.

I would like to know if you can help me where I can find technical specifications of video hardware requirements.

Because I want to integrate it with RIOT (https://riot.im/).

I would appreciate a lot of your support to find specific documentation for video in the aforementioned application.

best regards

What do you mean by video hardware requirements?


yes, requirements if you need to implement jitsu on a linux server for a scenario of 10,000, 2,000 or 5,000 users
That supports videoconferencing and does not saturate the server and fall for the same.
I have read from some the experience that 4GB RAM and 2 cors but, there is no official product documentation that says this is recommended for small or large companies.
Can you help me with your experience, or if you have even better documentation?

There is no official recommendation, we only give an example how we run meet.jit.si, signalling node and jvb instances we use are c5.xlarge aws instances.