Jisti Conference with Up To 100 Participant

Hi,

Today we tried conference, which went up to 94 participant on our local Jitsi deployment. Audio was very much perfect but video was not up to the mark. below are some scenario we faced.

a) Some participant were continuously disconnecting.(Unable to connect. Reloading Page coming)
b) Some are complaining that there chrome browser gone to 100% and crashed.
c) No one is able to view all participant as maximum 35 participant can be seen on screen and after that there is no scroll where we can scroll to see rest of participant.
d) everyone complaining that there bandwidth consumption is very high.

Is there any mechanism to achieve conference up to 100 participant?
Is there any mechanism to control video quality for some participant.(means maintaining HD quality for some important participants/presenters(more than 1) and reduce video quality of rest to achieve video for all participants) ?

Any idea ?

1 Like

We are working on increasing the limits, for example right now meet.jit.si is limited to 75 participants at most.
One thing is to have multiple bridges so you can spread the load of a big conference, you need octo enabled. Also we are working on improving UI. We also advice of using and lastN for such situations so you limit the numbers of streams sent to clients.

2 Likes

Hi @damencho,

Thanks for the quick response as always.

I am going through lastN documentation, The video of “N” Dominant Speaker will be send to everyone for better performance which definitely help.

I have one question : Dominant Speaker is identified by Audio Level. Can we configure Dominant Speaker to be identified on the basis of
a) Audio level
b) on the basis of some other flag.(Say if he is a moderator. Every Moderator can be considered as dominant speaker does matter he speak a single word)
c) (The Participant who raise the hand) ?

Is any custom logic can do it ?

Nope, there is no such feature.

Hi @damencho,

So is it possible with octo and multiple jvb components to support single conference with more number of users.

Suppose if I need to host a conference with a capacity of 200 users, Can I configure octo according to https://github.com/jitsi/jitsi-videobridge/blob/master/doc/octo.md and setup the conference.And add multiple jvb components to further scale up and host more participants in the call.
Also do you have any documentation on setting up region based bridge selection.

Thank you,
Nithin

How do we need to configure it? Currently the RegionBased split seems to spread people over the wolrd during a ~60 people EU centric call with some US and ZA participants.
I would expect to use the other eu-central-1 instances instead of a world-wide spread of participants.

@damencho, is there any update regarding the configuration/architecture of server to host a conference up to 100 participants? if no, is there any estimsated time we can expect to have it?
Thanx in advance :heart:

Well we are working on that, the work is mostly client. But there are some bridge updates.
So about architecture, nothing is changing all you need is octo enabled so you can have one conference in multiple bridges.

So right now it is possible to create meetings up to 100 participant with those updates and updated config?

I have not a great idea about OCTO (plz correct me if I am wrong and where should I look to upgrade from prev docker config) but my rough idea is other components are nearly same and we are just multiplying the video bridge instances to balance the load on server? but wasn’t it already in jitsi in past for auto scaling? but that actually meant for increasing the possible number of conferences… not for increasing the number of participants in 1 conference, am I right?

Then what/where is the main differences (in OCTO) that enables jitsi to create rooms with nearly 100 participants which wasn’t possible in past from high level view?
Thanx in advance :heart:

You distribute the load between several jvbs, otherwise for 100 person call the stream number is so big that you overload the single bridge that is hosting the conference.
These are 100 incoming streams and 100*99 outgoing, so in total 10000 streams.

but this autoscaling feature also existed as “autoscaling” by scaling jvb instance where jvb get scaled when new load arrives that I saw in tutorial but also then it was not possible to create rooms may be not more than with 30 participants?
then this OCTO feature enables to handle 10,000 streams by same thing by scaling jvb… this area I dont get it :disappointed_relieved: what is it doing actually…
then how the client load is being handled? 100 incoming streams for each participant? I am just curious about the medium level view…
and thanx for the reply

If you don’t have octo one conference uses one bridge only. So that bridge needs to handle all 10k streams, which is too much.
If you have octo you can spread the conference on several bridges and so are the 10k streams.

1 Like

@damencho
I have faced on several issues, can you please help me to sort out these things?
I did the Jitsi normal installation of a server. Then did the setup of two additional JVBs, two servers, and connected to the prosody.
Then I enabled OCTO.
I need to know is, in “RegionBasedBridgeSelectionStrategy”, if we enable OCTO according to this way, are we supposed to do this work by only using the AWS?
Another thing is, how does the JVB decide jitsi, according to how the participants should be connected?
Because in one article I saw that, it needs to define JVB’s region in Jitsi Meet’s config.js.
I need to know, how to do that dynamically also.

If you don’t want to do the regional part and just split a conference in several jvbs. Make all bridges report same region, and same and for the participants.

I think implementing VP9 also help increasing the participant count on each videobridge. cmiiw

yeah I get it for server side… but what about client side? as jitsi architecture is highly bandwidth dependent how can one client handle 99 incoming streams? just that clients bandwidth needs to be able to receive those streams or there is some optimization for this case from server side?
Thanx in advance :heart:

Yes, lastN is called. Where when configured you say how many of the last active N streams to be sent to clients.

Oh thats mean this outgoing streams from server is limited?
for client side this is optimal may be for 20-30?

@damencho
Thank you very much for the quick reply. I have a requirement like this which is, in AWS we can use separate regions. Not like that, if I need to deploy in like US and then do the load balance among them, is it possible?. It measn there are 54 regison in US. How we can do like this.