From: Radoslav Vasilev (deckland@gmail.com)
Date: Sun Sep 17 2006 - 13:26:09 ART
Oook,
On your second point - was checking with Wendell Odom's DQOS book as i
remembered that CB Policing is exactly like CAR in regard to the Be value -
both are the sum of Bc and the actual Be...
Any idea how can this be check guys?
On 9/17/06, Petr Lapukhov <petr@internetworkexpert.com> wrote:
>
> Yep, you are correct here.
>
> There are two main differences between MQC policer and CAR.
>
> 1) Exceed action is "pseudo-random" with CAR (compound/actual debt
> algorithm) to simulate RED behavior. With MQC there is three-color
> single-rate
> marker algorithm. There is no "violate" action with CAR and "exceed"
> action
> is "deterministic".
>
> 2) "Be" value with CAR is used to calculate "margin = Be-Bc", which is
> used
> for debt computation. With MQC, it is the actual Excess bucket value.
> [You didn't account that fact in your configuration]
>
> Therefore, strictly speaking, there is no strightforward way to convert
> CAR configuration into MQC. A most close approximation would be:
>
> rate-limit input <..> 256000 4000 8000 conform-action transmit
>
> exceed-action set-prec-transmit 0
>
> ==>
>
>
> police cir 256000 bc 4000 be 4000
> conform-action transmit
> exceed-action transmit
> violate-action set-prec-transmit 0
>
> HTH
>
> 2006/9/17, Radoslav Vasilev <deckland@gmail.com>:
> >
> > Hi Group,
>
> While re-testing some QoS sections, I came accross this lab task, which
> solution I consider partially wrong.
> This will revive an old discussion about exactly this task (CAR to MQC
> configuration transition).
>
> I won't copy the whole CAR configuration, as I have problems with the
> following part of it only:
>
> existing rate-limit:
> [...]
> rate-limit input access-group 102 256000 4000 8000 conform-action transmit
>
> exceed-action set-prec-transmit 0
> access-list 102 permit tcp any any
>
> This has to become a CB Policing congiguration with the IEWB solution
> being:
> policy-map MQC-CAR
> [...]
> class TCP
> police cir 256000 bc 4000 be 8000
> conform-action transmit
> exceed-action set-prec-transmit 0
>
> I consider the solution incorrect as:
> CAR and CB Policing have multiple differences, one of which is that CAR
> has
> two actions (conform, exceed) and CBP has thee (conform, exceed, violate).
>
> This is true when Be>0, which is the case in this task.
>
> Now, what CAR achieves with the configuration above is that traffic will
> still conform even when additional tokens should be borrowed from
> Bucket2(Be
> bucket). In case when Be is used, the packets Precende won't be override.
>
> With the proposed IEWB MQC solution, the packet's IP Precedence will be
> override when Be extra tokes have to be used (bursts of traffic when Be
> can
> cover). Therefore the suggested solution in my opinion is not entirely
> correct.
>
> My solution:
>
> class TCP
> police cir 256000 bc 4000 be 8000
> conform-action transmit
> exceed-action transmit
> violate-action set-prec-transmit 0
>
>
> Could you guys comment the solutions?
>
> Rado
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
>
>
> --
> Petr Lapukhov, CCIE #16379
> petr@internetworkexpert.com
>
> Internetwork Expert, Inc.
> http://www.InternetworkExpert.com
> Toll Free: 877-224-8987
> Outside US: 775-826-4344
This archive was generated by hypermail 2.1.4 : Sun Oct 01 2006 - 16:55:40 ART