Maximum participant in 2.0 version

[Sergey_P_1]

Good day!
 

You wrote me: 
> Hundreds of participants can be held on MCU simultaneously. We are still working on that exact number. Stay tuned!

How many participants does support 2.0 version?

Is it possible to increase number of participant  that can be held on MCU till 500-1000 ?

[Chunbo_H_Intel1]

Hi Sergey,

Thanks a lot for your continuous attention on our product!

Here is the thing. Since we just released our latest 2.0 with major architectural refinement, we are now working on the new benchmark data, such as the maximum participant number, to make it more accurate. We will release the data later, and will definitely let you know when it's ready.

[Artem_M_]

I have a question assotiated with maximum number of participants:

Sometimes(one of 8) when we execute ConferenceClient.join(...)
It doesn't connect to server.

but wen we do reconnect(get tocken and join again) - it connects.
On what it is depends on?
May by bad connection or ping? Or it can depends of turn-stun servers(we use default).

What is minimal brandwidth need for good connection?

[Chunbo_H_Intel1]

Hi Artem,

Can you share the deployment topology you're using now? Like, do you deploy the MCU server on public network, or behind a firewall?

Regarding the reconnect from client side, I think the reason is that we provide recovery mechanism on MCU server against system/network failure.

When you mentioned "It doesn't connect to server", do you mean the sample application server we provide, or the MCU server itself? I would suggest you upload the client output(like from the Chrome console) and server logs (Most of the server failures appear there), which can help us understand your user scenario and diagnose the root cause...

And, minimal bandwidth varies depending on the audio/video codec, video resolution, bitrate, e.t.c. Are you mainly using VGA with VP8 now?

[Artem_M_]

1) I use server with public Ip. No firewall
2) We worked with sample application. But the problem was with erizo controller. Url like https://*****8:8080/socket.io/1/?t=1422292671724 - beginning loading without any response. So conference couldn't join.
It's not part of sample application. Because tocken recieved ok, but connection to erizo fails.
3) We use default properties with vga(config.erizo.openh264Enabled = false - it's VP8)

I don't know what it depends on. Because sometimes it happens frequently and sometimes very seldom

[Chunbo_H_Intel1]

Yes, it seems to be the sock.io (8080) connection issue.

Since you have an MCU server on the public network, how do you deploy the client application during the test of erizo connection failure? Is it behind the firewall, or with public IP as well?

We will try to reproduce this on our side.

[Artem_M_]

Our client application isiirewalls. also in public network. We don't use firewall.

It will be hard to reproduce. After we began used more powerfull server and disabled mixind - I cant't catch that problem on our enviroment. 

But if I face it again - I'l try to get all logs and all that can help you. May be such problem can be according to highly loaded server. Don't know exact.

[Chunbo_H_Intel1]

Thanks for the input. I think we might meet the similar issue when we run scalability test on MCU server continuously. Will let you know more if it comes.

[Artem_M_]

One question not about scalability, but important for us.

When we request tocken - we set role: publisher o viewer.

Permitions for any role we can find in property files.

But can we connect all users as publishers(only some of them will realy publish - other the most time will only view).

Is there any restrictions for that on server side? Or checking of maximum publishers executes only wen someone starts publish?

[Chunbo_H_Intel1]

On the server side, permissions like {"publish": true, "subscribe": true, "record": true} will be respected regarding publish API, subscribe API and record API which will be available in the coming release.