QOS or more bandwidth

Pete Kruckenberg pete at kruckenberg.com
Tue May 29 19:19:59 UTC 2001

On Tue, 29 May 2001, Kavi, Prabhu wrote:

>> Voice-over-IP benefits from statistical multiplexing as
>> much, if not moreso, than any other application. A
>> toll-quality voice call runs at ~5-6kbps (factoring silence
>> suppression and RTP header compression) vs. 8kbps across
>> compressed TDM.
> Do you mean G.723.1? Let's put aside that some people do not
> consider G.723.1 to be toll-quality.  Ignoring that, I would
> call your 5-6 kbps number to be compression rather than
> statistical multiplexing.  The voice packets still come out
> periodically every 30ms or so.  The only significant statistical
> multiplexing you get with voice is due to silence suppression.
> Also, how do you actually achieve 5-6 kbps when you
> consider the IP overhead?

This was information that came back from one of our test
networks, that I didn't participate in personally, so I may
be speaking apples to oranges.

I have done the theoretical numbers, an 8kbps raw stream
takes ~21kbps without IP header compression (mostly because
of the packet-per-30ms requirement). The numbers that came
back from the test showed with RTP header compression (takes
care of some of the IP header over head, at the cost of some
CPU state-tracking) less than 8kbps. I was surprised. Maybe
I should look into it further, but I had heard similar
numbers in casual conversation with other people.

Curious what others would consider 'typical' bit rate for a
"toll-quality" (as far as the users are concerned) VoIP
session, incorporating the various compressions and
optimizations that are available.


More information about the NANOG mailing list