[c-nsp] Question about ip rtp header-compression

[Oliver Boehmer (oboehmer)]

Ziv Leyes <> wrote on Tuesday, February 05, 2008 11:50 AM:

> Hi,
> I have a problem I can't figure out myself.
> I have two 7206VXR connected between them with serial interface over
> satellite. The bandwidth is 2Mb (clockrate of the controller shows
> 2047212 on both sides)  
> This link is exclusively used for VoIP, and the ip rtp header
> compression is activated on both sides with a very nice successful
> statistics, such as 99% hit ratio and around 2,50-3,00 efficiency
> improvement factor.   
> The customer still complains about several VoIP packet loss, and in
> the line graphs you can see the line never exceeds the 1500-1600 Kb. 
> I wanted to try to improve the rtp traffic so I thought about using
> "ip rtp priority 13000 16383 2000" just to give it a try, but it gave
> an error saying "IP RTP: Not enough bandwidth: available 1500 needed
> 2000"   
> Then I realized there is a hard limit of 1500 for the IP RTP, and I
> wonder, where is this limit coming from? 

by default, QoS, incl. "ip rtp priority" only uses 75% of the link
bandwidth. You can increase this using "max-reserved-bandwidth 100". I
wouldn't use 100%, otherwise you'll starve your routing traffic (if you
run routing over the link). 
You can also use CBWFQ/PQ queuing using MQC (policy-map, class-map,
etc.) to prioritize your traffic.

	oli