[c-nsp] configure 876 with qos

[Adrian Chadd]

On Mon, Jul 16, 2007, Rodney Dunn wrote:

> > Why would you police/bandwidth on inbound data? You've already
> > received it.
> 
> I used to argue that same point. But I've since backed off
> of it a bit. The reason is a deployment scenario where a 
> user (customer) connects but has multiple egress interfaces
> towards the core/destination. You want to provide them some
> form of per remote shaping/policing. Meaning you want to provide
> the same service towards the remote irregardless of the egress
> path to get there. So if you can apply the policy to the interface
> for that specific user on ingress (their attachment circuit) it
> does make the concept of shaping/queueing relevant.

Oh, I'd understand that for something bigger than an 876.
Thats why I then use marking on incoming and policers/aggregate
policers on egress interfaces; trying to get all the QoS features
needed on the (lowish-end) L3-switch platforms gets tricky.




Adrian