[c-nsp] 7200 DSCP-to-EXP Mapping

[Mark Tinka]

On Monday, September 19, 2011 05:35:11 PM ar wrote:

> Thanks. I dont want to apply policy on all ingress
> interface of the PE to remark traffic. That's why I want
> to be sure on this TOS reflection..

Well, I always think it's best to mark on ingress because 
that's the closest interface toward downstream customers. in 
deployments where you have a router that is both a P and PE 
device (such as in a Metro-E ring, e.t.c.), it can be 
difficult to differentiate transit (P) from terminating (PE) 
traffic.

This is why I've never liked Juniper's egress 
rewriting/remarking policy in earlier line cards, as it 
really limits the kinds of topologies you can have with 
certain platforms. Suffice it to say, this is now changing 
in later line cards, but leaves much to be desired.

Anyway...

> If this is the case, is it safe enough to just do a
> policy-map on the  PE-to-P interface matching for the
> reflected EXP bits?

In theory, yes, as the first 3 bites of the DSCP value will 
be mapped to the EXP bit.

But in order to ensure that you get the right mapping from 
DSCP to EXP, you may end up marking DSCP traffic on ingress. 
If you do that, you might as well just mark EXP bits on 
ingress :-).

Otherwise, I can't promise that the values you're matching 
upstream on the PE-P link will always mirror what was set as 
the packets came into the PE router in the first place.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20110919/0a6bb5e1/attachment.pgp>