[Asterisk-Dev] IAXtel Status

[Daniel Pocock]

Mark Spencer wrote:

> The Asterisk+SER combination does seem to be very commonly deployed in 
> SIP environments although I haven't messed with it myself.  Obviously 
> SER does not support IAX and in any case, I think we should be trying 
> to solve this problem with Asterisk and really taking the time to 
> explore what Asterisk's performance limitations are and why.
>
> Mark
>

The big difference between SIP and IAX is that SIP handles just the 
signalling, and the audio (and video) is sent in separate RTP streams.  
SER doesn't handle the RTP streams, so naturally SER is much more 
scalable.  Therefore, I'm not sure that SER supporting IAX would be 
appropriate.  There is the possibility of creating a pure IAX equivalent 
of SER (perhaps a cut down Asterisk that just does IAX, no transcoding 
or media processing, etc).

Furthermore, SER doesn't have to worry about timing - it just forwards 
requests as soon as they arrive.  Asterisk does have to worry about 
timing, whereas a pure IAX-IAX switch could possibly forget about timing 
and leave it to the end-points to worry about.

IAX combines the media streams and signalling into the same packets (the 
RTP packets NEVER go through SER).  The benefits to this are a) trunking 
and b) easier NAT traversal.  I think the downside is that having all 
this extra stuff in the protocol will inevitably put a limit to the 
number of live sessions the NIC can support concurrently.

This doesn't mean that IAX is somehow bad and SIP/SER is good - it just 
means they fill very different needs in the marketplace.  Eg companies 
wanting to trunk calls between remote offices over satellite links would 
find IAX much more cost efficient than SIP.