CAR & POLICE - Better impact on CPU ?

Unanswered Question
Mar 8th, 2010

  Hi All,

I have a requirement to rate limit one of the customer to 40Mbps. Unfortunately I have to do it in Cisco router. The customer is landed on POS interface with encapsulation configured as FR. Customer has got 2 services so 2 different DLCI are used. In one DLCI, 40Mbps CAR is configured. Whenever there is a flap in the POS interface the router has very high CPU. I guess configuring CAR makes the router to do a process switch of every packet it receives.

If policing is configured rather than CAR does it reduce the load on CPU. Because class-map and policy-map works on CEF and it reduces load on CPU.

Would like to know:

1. When CAR is configured, router does Process switching or still does CEF. How about the load on CPU.

2. If policing is configured, router does CEF. But how about the load on CPU. Will it be the same like CAR or somewhat better.

thanks in advance


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Giuseppe Larosa Mon, 03/08/2010 - 12:49

Hello Arun,

>> I guess configuring CAR makes the router to do a process switch of every  packet it receives.

no CAR requires CEF and uses CEF, modern class based policing requires CEF too.


CAR or VIP-distributed CAR can be configured on an interface or  subinterface. However, CAR and VIP-distributed CAR are not supported on  the following interfaces:

Fast EtherChannel



Any interface that does not support  Cisco Express Forwarding (CEF)

use sh proc cpu sorted during flapping to see what processes use more resources.

if I understand correctly you are using Frame Relay encapsulation over an OC-3 POS I remember it was not possible to use FR over POS, what platform are you using?

Years ago the fastest FR link was a T3 link as far as I know.

or is this an ATM OC-3 interface?

Are you limiting inbound or outbound with CAR?

if outbound you could consider shaping as suggested by Paolo

Hope to help



This Discussion

Related Content