Showing results for 
Search instead for 
Did you mean: 

Multilink PPP fragment size


I've got an hub and spoke topology with all interfaces running MLPPP. I've got Video over IP on the spoke sites (1720 IOS 12.2.7b) and the core site (3640 IOS 12.1.5 T8). The WAN links are all 640k and all routers run cef switching. There is a service policy to prioritise the video on the edge sites and the core by means of a class map matching the IP addresses of the video devices i.e any source address configured in the access list is the address of a video device.

Okay, no worries.

The video works fine between the hub and the core. However, I get breakup if the video tandems across the core. In this situation, will the core 3640 check the access list for tandem service policy connections ?

Furthermore, the video packets are all 1500 bytes long. Is there any point in setting a fragment delay under the multilink PPP since the video stream is made up of the maximum allowable ethernet packet size. What's the point in breaking up these packets since I wouldn't get any improvement in perfromance ?

Am I thinking along the right lines here ? I can see no bugs with 12.1.5t8 regarding my problem specifically though these are numerous QOS issues with that release. Should I upgrade to 12.2.7b on the core as well ?

Any help is much appreciated. Thanks. Steve, England.


Re: Multilink PPP fragment size

Fragmentation is recommended on any link less then 768K, so you are right on the border. However, it's large data fragments that need to be fragmented, and not voice or video packets. Furthermore, anything that goes through the LLQ priority queue is not subject to fragmentation. Rather it is interleaved between the data fragments.

It's impossible to troubleshoot this one without much more info. Do your ACL's everywhere check for the IP address of every video device? They'd have to if you want to be able to send video between any 2 terminals.


Re: Multilink PPP fragment size

Thankyou for the input,

Yes the ACL's check for the IP adresses on each site. In addition, the ACL at the core "tandem site" are configured to match the source IP address of all of the sites to allow priorititisation to occur for tandem video connections. I can see the incrementing packets in the "show policy map interface multilink 1" command. The configs must be correct regarding the multilink interfaces or else it would not work on a point to point basis. Why then does it break up on tandem connections ?

Thanks if you can give me any further input.