Configuring CallMgr bandwidth in a location w/ RTP compression

[Robert Milanovich]

I have a number of sites which are connected via Frame, using G.729, and RTP header compression.

According to Cisco docs, each G729 stream uses approx 26bkps with two streams being used per call for a total of 52kbps.

With header compression, each stream should drop to approx 15kbps so the total call bandwidth should be approx 30kbps per call.

When I set the location bandwidth setting, should I be accounting for the compression since CallMgr doesn't know anything about the compression and the savings?

For example on a 256k pipe:

I can get 4 uncompressed header calls through

Using compression, I can do 8 calls

So, when I enter the location bandwidth settings, would I want to enter 256 for it or 512k?

My reasoning is that I would want to do 512k so that I can get the full amount of calls (eight total with header compression).... where if I did 256, then I could only do 4 calls since CallMgr is assuming that I am using uncompressed headers.

[dugrant]

You are right, CM doesnt know what the actual BW being used is. Depending on the L2 encapsulation, and compression settings the bw can vary.

CM assumes that a 729 call is 24k, and an 711 call is 80k. It is up to you to determine how many calls CM should allow across the WAN.

If you have configured your WAN to support 4 calls, then in CM you will enter 96 (4x24k) into the location configuration.

Also remember that the location is assumed by CM to be the central site, so all devices in the central site should be in the location.

[Robert Milanovich]

Just to verify...

A single call uses two streams which means in a G711 situation it would use 160k total...

Correct?

[pbarman]

A single uses two udp streams, but at one time only one party is talking, so in effect it averages to one stream only. And for one call, you need to specify 80 kbps, and not 160 kbps.