Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 
New Member

Mapping Model in MPLS VPNs


Based on paper titled "L3 MPLS VPN Enterprise Consumer Guide" page 52, figure 44. (

1) The figure discards the "streaming video" and "bulk data" traffics within the mapping process. Why? What happens with these traffics? Both traffics are discarded or simply they need to be mapped to "Best Effort"? Please explain.

2)In the same figure, "Interactive Video" is mapped to "Realtime" SP class with "Voice" traffic. Is this "Interactive Video" traffic always no TCP-based? If the opposite is true, why is it mixing TCP & UDP over the same "Realtime" class?


Re: Mapping Model in MPLS VPNs

1. What the document is saying here is that this 3-class PE model does not support streaming-video or bulk-data therefore the CE should not be submitting this traffic at all. It's not saying that the PE will discard it. The PE will accept anything you give it - it's just that performance of your other traffic will suffer so it's best not to submit this traffic at all.

2. Most interactive traffic is UDP-based. There is no real need for TCP's reliability in such scenarios. Therefore, both types of traffic in the real-time class are UDP-based.

Hope that helps - pls rate the post if it does.


New Member

Re: Mapping Model in MPLS VPNs

Hi Paresh:

According to this article, streaming servers by default use UDP protocol, so I don´t understand why the Cisco paper classifies "streaming video" as a TCP sesion. Maybe is it linking streaming video with a HTTP sesion? Please explain.


Re: Mapping Model in MPLS VPNs


That articles mentions that these protocols tend to use transport-layer protocols such as UDP and RTSP. That is true but there are a lot of different streaming protocols around and some of them do use TCP. In fact, even RTSP supports the use of TCP. And you can also stream via HTTP (Windows Media supports this, for example).

So you see, there can be a mix of TCP and UDP traffic here.

The other, more critical, reason for not mixing interactive-traffic with streaming (one-way) traffic is the drastically different jitter/latency requirements for the two. Streaming traffic will easily sustain latency in the order of seconds and jitter is not even a problem. Whereas interactive traffic will not. That is why you should not mix the two.

Hope that helps - pls rate the post if it does.


CreatePlease to create content