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

Attention: The Community will be in read-only mode on 12/14/2017 from 12:00 am pacific to 11:30 am.

During this time you will only be able to see content. Other interactions such as posting, replying to questions, or marking content as helpful will be disabled for few hours.

We apologize for the inconvenience while we perform important updates to the Community.

New Member

Class-Based Policing - Recommended Burst Sizes


Can anyone clarify what the recommended burst vaules are for Class-Based Policing? The documentation for the rate-limit command suggests a normal burst of CIR [BPS] * (1 byte)/(8 bits) * 1.5 seconds and a extended burst of 2 * normal burst. For example:

rate-limit 1000000 187500 375000

The documentation for the police command is not as helpful and only provides the following example:

police 8000 1000 1000 conform-action transmit exceed-action set-qos-transmit 1 violate-action drop

If the burst size is not specified with Class-Based Policing the IOS defaults to a Bc of CIR/32 (250ms Tc) which is very different to the 1.5 seconds value Cisco recommends for the rate-limit command.

The documentation for the rate-limit command also states “Setting the extended burst value equal to the normal burst value excludes the extended burst capability.” I take it this is not the case with Class-Based Policing as the Cisco example has 1000 bytes for both normal and excess burst sizes?

If anyone can explain the differences between Class-Based Policing and Committed Access Rate burst values it would be appreciated.



Hall of Fame Super Silver

Re: Class-Based Policing - Recommended Burst Sizes

Hello Paul,

CAR has only two possible states:

conforming and exceeding.

Class based policing is different because it has three possible state for a policed packet:

conforming, exceeding, violating.

both features use a token bucket algorythm to "count" traffic against the configured rates.

Class based policing uses Be when Be <> 0, token rate fills first Bc and then spills down to Be until it fills it.

In this way unused tokens in previous intervals can be accumulated up to Bc+Be and are available at next interval.

CAR implements its token bucket differently.

in CAR extended burst size has effect if Be > Bc and actually what counts is Be / Bc ratio at least this was our impression when we tested CAR some (well 10 ) years ago

A useful document is enterprise QoS Solution reference network design

Hope to help


CreatePlease to create content