When to MTP, when to transcode ?

That was a question I saw on NetPro today (paraphrased slightly), which made me think of something I'm currently working on.

I'm currently involved in the implementation of a new Audio Conferencing system, which involves Avaya conferencing connected to Cisco IPT. This involves SIP trunking between the IPT platform and the conferencing system. This brings into play the usual "how do I get my DTMF to work?" question.

Normally, I'd just create an MTP resource on an IOS gateway, add it into a Media Resource Group and apply that to the SIP trunk. This works fine when calling from an IP phone to the conferencing system, but when an MGCP gateway gets involved, its time for some interesting behaviour - if you type your DMTF quickly, digits get dropped - not all though. If I use a Transcoder (same gateway etc), it works fine.

Now a transcoder by default uses two DSP sessions per call (even when going from G711 to G711) as the default configuration sets up allowed codecs to include G711 and G729. So take away the G729 as an allowable codec and its down to one DSP session per call, just like an MTP.

I'm guessing this could be a bug with UCM 6.1.2 and the IOS version I'm using, but I have a project to complete, so I'll have to come back to this at a later time as the workaround is suitable

So to answer the initial question, I guess the answer might be, use a transcoder all the time and control your codecs to optimise your DSP usage - unless you are doing a certain lab exam!