What is the implementation principle of JVB(version:5142) load balancing? Is there any relevant documentation to explain?

What is the implementation principle of JVB(version:5142) load balancing? Is there any relevant documentation to explain? We want secondary development, we want to understand how it works. Is there documentation about the implementation logic in JVB source code stress_level?

FTR JVB is not doing load balancing by itself; it’s Jicofo that is introducing users in the rooms and is basing its decisions on stress_level indicator( calculated indeed by JVB), to create a new room on a JVB or another.
stress_level seems to be documented only in the Jvb code. From what I see I’d say it’s based on traffic level information given by RTCP.

  1. Is it possible to fetch the stress-level via API?

@bbaldino @gpatel-fr

it seems to be generated to be included in Json returned by Colibri, so my guess is it’s the only way.

1 Like

Thanks @gpatel-fr
I’ll try it tomorrow. it’s 1:00 AM here. time to :sleeping:

137/5000
Thank you for your reply. What plugins do you use to calculate RTCP and JVB? JVB statistics of the total number, but I can’t understand incomingPacketStreamStats. PacketRate and outgoingPacketStreamStats packetRate sources. Can you give me an example? Very grateful

sorry, I just don’t know. My best guess is that this is a sum of the bandwidth estimations sent by the endpoints, that is, the Jitsi clients (no idea where this happens in the clients, if it’s in lib-jitsi-meet or at a deeper level in the browser, that is, a browser library of course) anyway see REMB