Decoding the codec

[nash]

The best codec to use has to be G711. I don’t know why anyone would even bother with G729 especially now bandwidth isn't an issue. Can anyone out there give me any good reason why I would use G729 on my system?

[gsgleaso]

The best codec to use has to be G711. I don’t know why anyone would even bother with G729 especially now bandwidth isn't an issue. Can anyone out there give me any good reason why I would use G729 on my system?

Internal to your own LAN, there is no reason to use an inferior codec.

Traversing WAN links, however, is another story. Bandwidth is expensive and limited, and a compressed codec can allow several more calls than an uncompressed codec.

[nash]

Is bandwidth such an issue nowadays? Is it still a problem for us that we need to think about compression on a WAN link?

[roberto]

Given the broad array of codecs and vendors today, it is hard to know what you will need to use when establishing a call. That's why systems now allow you to have an "allowable" and "prioritized" codec list. One enterprise may decide that G711 should be the first thing to try to negotiate when dialing local extensions within the WAN. It may however choose a G729 when sending calls between locations since it has to cross WAN boundaries and trunking becomes a requirement.

Bandwidth becomes a concern if you have to traverse unmanaged networks (public internet, for example) where DSCP coding is not an option and you are left at the mercy of "best effort" routing. In such cases, minimizing the amount of data flow between endpoints becomes a necessity, since less data means less chance of packet loss.

Would I rather have a lower quality, but coherent, call or a high quality but stuttery call? I'd pick coherency first.

Then adding cost to the mix. In this day of bandwidth caps, data transfers can become costly. In developing countries such as the BRIC, the communications infrastructure is still evolving and transfer costs are high given the high demand and low supply economic drivers.

<With my head in the clouds>
If bandwidth was unlimited and latency negligible, and I controlled the end-to-end network entirely, would I deploy G711? Perhaps at first (to make sure it all works. ). But afterwards, I'd want to maximize the voice quality with the highest possible codecs I could find (even royalty ones). Why not...in an ideal world.

<Looking down, and slowly descending to Earth>
In today's world, however, I'd measure the bandwidth cost between endpoints and then optimize the codec lineup to save me the most $$$, tailoring a full-scale endpoint matrix based on the traffic patterns, time-of-day usage, and calling habits of each endpoint. I would then overlay the whole network with dynamic contextual intelligence that would self-optimize the codecs over time so that I was always using the best possible codec with the least amount of cost while keeping the calls coherent.

<Touchdown. Back to reality>
Hmmm...perhaps G.711 is the only thing we need.