[Asterisk-video] MCU architecture

[Sergio Garcia Murillo]

Hi everyone,

Currently the mcu solution has two main components, the VideoMixer and 
the mcuWeb.

The videomixer component handles ONLY media, i.e. it receives rtp, 
unpack audio and video, performs audio/mixing video, encoding, packing 
and rtp sending. It is completely controlled by a xmlrpc interface and 
has no service logic at all.

The current xmlprc api has the following methods:

-Create/Destroy conference
-Add/Remove participant to conference
-Set conference parameters like video size and number and distribution 
of participants on screen
-Set audio/video send/receive ports per participant
-Set audio/video codec and parameters (size,fps) etc per participant
-(Un)Mute participant
-Set conference mosaic positions: lock slot, assign slot to participant, 
etc..
-Add watch only participant to conference (experimental: flash video 
broadcasting in web)

It currently supports only h263p but should not be too difficult to add 
support to h264. The rest of the functionalities are completed (except 
flash support) and only a bit of testing is needed. In the near future I 
would like to convert the videomixer in a MediaServer, been able not 
only to offer multivideo conference services, but also transcoding, 
flash casting, etc..

As I said before everything is controlled by an xmlrpc api, so a 
component handling the service logic and signalling is needed. That 
component could app_conference and the confiance project did integrate 
the video mixer as an external unit.

I decided to implement it as a complete external unit from Asterisk. 
Why? I think it was easier quicker and easier to develop, avoid the 
monolithic and sometimes obscure architecture of asterisk and could 
provide much more functionalities. And the chosen technology was.. java 
(I feel a great disturbance in the Force, as if millions of voices 
suddenly cried out in terror and were suddenly silenced).

Yes, Java, using the Sailfin Sip Application Server 
(https://sailfin.dev.java.net/) which allows to create an application 
that handles SIP and HTTP request (an application like click to dial is 
just a few lines of code 
http://wiki.glassfish.java.net/Wiki.jsp?page=SipClickToDialExample2). If 
you start with your prejudges about java, speed and show on, just think 
that it is a telco grade Sun and Ericsson development.

The mcuWeb component implement the service logic, handles all the SIP 
signalling (receiving invite request from asterisk), controls the Video 
Mixer with the xmlrpc and offers a WEB UI to manage the conferences.

This part is also fully functional, but I think that is where more work 
is needed in order to customize the service with the functionalities 
needed by the customers. In particular questions like the following need 
to be answered:

- Is it required to create the conference before the user calls? or it 
get created when it calls in?
- Are there private conferences? How are the participants allowed to get 
in, by password or by invite only?
- Is there always a default public room?
- etc...

Any thoughts are welcome

Best regards
Sergio