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

policing and excess burst sizes

Cisco docs state for cb-policing...

Defaults: Bc = cir/32, Be = Bc (burst disabled).

Recommended: Bc=cir x 1.5, Be = Bc x 2

1. For both CAR and CB-Policing is it true that...

If Bc = Be, excess burst is disabled. Be must be greater than Bc for excess bursting capability.

2. If so, I'm missing why we can't use the Be if it's the same as Bc.

If Bc = 4K, and Be = 4K, in time interval 1 (t1) if I send 4K conforming, the Be should allow me to burst another 4K as excess. But I can send no excess if I'm reading it right. If that's just the way it is I can live with it.

3. If Bc = 4K, and Be = 6K, in time interval 1 (t1) if I send 4K conforming, does the Be allow me to burst another 6K (10K total in t1) or another 2K (6K in total) as excess?

4. In CAR the max-burst (Be) is configured, but no violate action is possible. Is there a second (excess) bucket used for regulating max-burst?

Thanks.

1 ACCEPTED SOLUTION

Accepted Solutions
Purple

Re: policing and excess burst sizes

Randy,

If you extrapolate the CAR recommendations to CB-Policing, you get:

Bc = Be = 1.5s at CIR

Strictly speaking, if your committed bucket has 2k in it, and the excess bucket has 2k in it and a 3k packet came long, it would violate.

Pls do remember to rate posts.

Paresh

7 REPLIES
Purple

Re: policing and excess burst sizes

Hi,

The recommendations you have stated seem to be for the older implementation of CB-Policing, where it was modeled on CAR. The current implementation is based on the following RFCs: RFC2697 & RFC2698

In answer to your questions:

1. For CAR, Be is inclusive of Bc. Therefore, if Bc = Be, there is no excess burst capability. For CB-Policing, Be does not include Bc, therefore, there is no excess burst capability when Be = 0. If Be > 0, then you have excess burst.

2. My respose to 1 answers this question, I hope.

3. See response to 1. For Car, you will get 2k of excess burst. For CB-Policing, you will get 6k of excess burst.

4. Wtih CAR, you only get a single bucket policer, so it's either conforming or exceeding. With CB-Policing, you have the option of a two-bucket policer, which allows you the additional violate-action.

Hope that helps - pls rate the post if it does.

Paresh

New Member

Re: policing and excess burst sizes

Thanks Paresh,

What are the recommended burst equations for cb-policing if not the same as for CAR (Bc=1.5xCIR, Be=2xBc)?

1. excellent

2. it did.

3. ok.

So Bc can equal Be.

If at any moment if Bc = 2K, Be = 2K and a packet comes in sized 3K, will the packet conform, exceed or violate?

4. ok.

Thanks again,

Randy.

Purple

Re: policing and excess burst sizes

Randy,

If you extrapolate the CAR recommendations to CB-Policing, you get:

Bc = Be = 1.5s at CIR

Strictly speaking, if your committed bucket has 2k in it, and the excess bucket has 2k in it and a 3k packet came long, it would violate.

Pls do remember to rate posts.

Paresh

New Member

Re: policing and excess burst sizes

Ok. 1.5 secs for each individually.

Thanks Paresh.

Is this an official recommendation from Cisco?

Do you have a link for it?

Thanks again,

Randy.

Purple

Re: policing and excess burst sizes

Hi Randy,

I have not seen that as an official recommendation. I was just trying to extrapolate the CAR recommendations to CB-Policing.

In fact, the recommendation from Cisco is probably 250ms, since that is what it defaults to when you don't explicitly specify the Bc when configuring the 'police' statement.

Hope that helps - pls rate the post if it does.

Paresh

New Member

Re: policing and excess burst sizes

Okay.

Yeah. .250 for cbwfq, .200 for LLQ.

If this is the official recommendation, it's way different than the recommendations from the CAR days. (1.5 seconds).

I've opened up a TAC case.

Don't understand why they've made this simple concept so hard to nail down.

Thanks, Randy.

Purple

Re: policing and excess burst sizes

Good luck getting a design recommendation from TAC !

My thinking of the disparity between CAR and CB-Policing recommendations is that CAR, being an older technology, was used in the days when it was likely that shapers were not very granular. Therefore, you had to have a course measurement interval. These days, shapers are more accurate so you can use a smaller measurement interval.

Pls do remember to rate posts.

Paresh

936
Views
0
Helpful
7
Replies
CreatePlease to create content