[c-nsp] CoPP Service Policy

Oliver Boehmer (oboehmer) oboehmer at cisco.com
Tue Aug 26 04:53:25 EDT 2008


Aftab Siddiqui <> wrote on Tuesday, August 26, 2008 10:29 AM:

> Dear All,
> 
> 
> I would like to know the difference in performance and implementation
> if I put the service-policy within the specified interface (e.g.
> gix/x) or with in the control-plane in globally.
> *First Option:*
> 
> Router(config)# *control-plane *
> 
> Router(config-cp)# *service-policy input* *service-policy-name *
> Router(config-cp)# *service-policy output* *service-policy-name *

there is no outbound conrol-plane policing.

> *Second Option:*
> 
> interface GigabitEthernetxx/yy
>  service-policy input *service-policy-name *
>  service-policy output *service-policy-name *
> 

What are you trying to achieve?

CoPP policy (first config) is processed only for traffic terminating on
the router, while the interface QoS policy is applied to all traffic
entering (or leaving) the respective interface. So the semantic is quite
different.
CoPP ensures that the aggregate traffic (from all interfaces) does not
exceed a certain rate, while the interface QoS policy is only looking at
the rate of this specific interface (assuming you want to use the policy
to rate-limit/police certain traffic to the box).

Another advantage of CoPP is the easy "filtering" as it is only applied
to traffic terminating on the router, so you usually don't need to match
on any possible destination address in an ACL/class-map.

Both policies are execued in hardware (there is an addtl. software
CoPP), no performance impact. 

You might want to look at http://tinyurl.com/5hew55 for more info about
CoPP..

	oli



More information about the cisco-nsp mailing list