Realizing conferences

Conferences can be realizes in many ways:

Centralized Server, Endpoint Server, or Distributed Conferencing
Media Server Component
Distributed Mixing
Cascaded Mixers
Transcoding media at a conference bridge (see RFC 5370)

Slide Notes

G. Camarillo, “The Session Initiation Protocol (SIP) Conference Bridge Transcoding Model”, Internet Request for Comments, RFC Editor, RFC 5370 (Proposed Standard), ISSN 2070-1721}, October 2008 http://www.rfc-editor.org/rfc/rfc5370.txt Links to an external site.

Transcript

[slide446] So how do you realize conferences? Well, either central servers, endpoint servers, or fully distributed. Now quite often, you have a media server. Why? Because we want to be able to have pre-recorded media, and we want to be able to play it whenever we want. How do we do the mixing? Probably in a distributed fashion, or in some cases with cascaded mixers, so that we actually build logically a tree of mixers, and we mix our media together. Why? Well, it turns out beyond a small number, a reasonably small number of media streams, this becomes computationally expensive to do. So therefore, what's the easy way to do it? You split the work up, and then you combine it together. However, what does it do? It decreases the delay. But of course, if we have a good fan-out, what do we get? We get basically a log x behavior in the number of streams. So it's not bad. And of course, you may have to do transcoding, because if we have one set of participants in Japan or North America, sending them PCM A-law coded audio won't do them any good. They need to have PCM µ-law coded audio if they're attached to a telephony network. So we need to convert.