[Asterisk-Users] TDMoE and codecs

[Steven Critchfield]

On Wed, 2003-09-10 at 13:51, James Sharp wrote:
> > On Wed, 2003-09-10 at 11:55, James Sharp wrote:
> >> If I have a system with 1 machine to handle incoming H.323 calls and
> >> then
> >> multiple machines to distribute them to T1 ports over TDMoE, where does
> >> the codec translation take place?  Does it take place in the master
> >> system
> >> or does it take place in each of the slave TDMoE systems?
> >>
> >> Also, any idea how many concurrent G.729 calls a system like this would
> >> handle?
> >
> > Not to make you sound stupid, but TDMoE is Time Division Multiplex over
> > Ethernet. TDM is the same as what is done on T1/E1 lines. TDM implies
> > either ulaw or alaw encoding depending on your location. So therefore
> > the codec translation occurs before the audio makes it to TDMoE.
> 
> That's what I figured.  I just wanted to make sure I had my brain in the
> right direction.
> 
> > You probably would rather do IAX2 trunking as it uses only the necessary
> > bandwidth needed for the in process calls. TDMoE will eat the same
> > amount of bandwidth no matter what is being transported. Add to this the
> > fact that just like real T1/E1 circuits, they can't be easily
> > dynamically added. They have to be provisioned and then deployed. IAX2
> > will just scale to the ends of your bandwidth and CPU without additional
> > setup.
> 
> So have one machine take the H.323 calls in and trunk them out to each of
> the slave machines via IAX2 trunking (something I've not used yet).  And
> this will move codec processing out to each of the slave machines?

If the remote ends can do the codec, then yes. If they can't deal with
the incoming codec, then it will be done at your h323 end point. The
benefit of IAX2 trunking is to cut down on your ethernet load and to
make expanding easier. Not to mention IAX2 is much better tested than
TDMoE.
-- 
Steven Critchfield  <critch at basesys.com>