From: darth router (darklordrouter@gmail.com)
Date: Tue Jul 03 2007 - 22:10:58 ART
Thanks guys,
I suppose shape peak was very hard to understand without some practical
application. I just called up my buddy, and he laid it out very simply:
shape peak is used when you have a potential sustained burst agreement with
the ISP, and your traffic is properly classified, so your priority traffic
never really gets dropped as long as it falls under the average (CIR) rate.
I kept thinking in my head, why the hell would anyone use shape peak... lol,
now it makes sense, you just have to think about it with classification. The
bucket analogy works great if you put the whole picture together. Yea, I
know, I am slow :)
DR
On 7/3/07, Mike Kraus (mikraus) <mikraus@cisco.com> wrote:
>
> Yes you can always send Be+Bc with shape peak in every Tc (it doesn't
> appear to look at previous Tc usage of Bc). So, to be more accurate, I
> guess it would be:
>
> 1) Shape Average: Within each Tc the router sends Bc worth of data, and
> buffers the rest for the next Tc.
> 2) Shape Average w/Be: Within each Tc the router sends up to Bc+Be of
> data, (if all the Bc was not used in previous Tcs) and buffers the rest
> for the next Tc.
> 3) Shape Peak: Within each Tc, the router sends up to Bc+Be of data,
> and buffers the rest for the next Tc.
> 4) Police: As Bc is exceeded, packets are dropped.
> 5) Police w/Be: As Bc exceeded, managed discard is used up till Be.
> After Be, packets are dropped.
>
> With that being said, and to answer your question, yes you could define
> an average shaper with a Be of zero and a Bc that is equal to a Bc+Be of
> a peak shaper and receive the same results. However, as soon as Be does
> not equal Bc, there would be no way to configure an average shaper to
> have the same behavior as a peak shaper.
>
> See excellent attached table, courtesy of Arul Kumar of Cisco.
>
>
>
> ________________________________
>
> From: darth router [mailto:darklordrouter@gmail.com]
> Sent: Tuesday, July 03, 2007 3:43 AM
> To: Narbik Kocharians
> Cc: Mike Kraus (mikraus); ccielab@groupstudy.com
> Subject: Re: Shape/Police Peak
>
>
> "2) Shape Peak: Within each Tc the router sends up to Bc+Be of data,
> (if
> all the Bc was not used in previous Tcs) and buffers the rest for the
> next Tc."
>
> I thought this is the same as shape average with Be defined. I was under
> the impression that shape peak will send Bc + Be every interval. e.g.
> the buckets fill up every interval (not dependent on Bc leftover to
> overflow into Be), and if the receiving end cannot handle it, packets
> will drop. Is this correct? If this is true, is there any difference
> from shaping peak, to just shaping average to a higher rate (same total
> value as you would have used with peak). Could you clarify? Thanks,
>
> DR
>
>
>
>
> On 7/3/07, Narbik Kocharians <narbikk@gmail.com> wrote:
>
> I totally agree with you, many people take this bucket concept
> too far. I think your generalization is pretty accurate.
>
>
> On 7/2/07, Mike Kraus (mikraus) <mikraus@cisco.com > wrote:
>
> Scott Morris had once made mention that the whole token
> bucket metaphor
> was induced by mind-expanding agents, and I am learning
> to agree. :)
>
> If I ignore the whole bucket concept:
>
> 1) Shape Average: Within each Tc the router sends Bc
> worth of data, and
> buffers the rest for the next Tc.
> 2) Shape Peak: Within each Tc the router sends up to
> Bc+Be of data, (if
> all the Bc was not used in previous Tcs) and buffers the
> rest for the
> next Tc.
> 3) Police: As Bc is exceeded, packets are dropped.
> 4) Police Peak: As Bc exceeded, managed discard is used
> up till Be.
> After Be, packets are dropped.
>
> By no means is the above all explicitly accurate, but it
> is a broad
> generalization that has kept me sane. If anyone has any
> suggestions/comments on the above, I would love to hear
> it!
>
> Here are a couple good pages I've found useful:
>
> http://www.cisco.com/en/US/tech/tk543/tk545/technologies_tech_note09186a
> <http://www.cisco.com/en/US/tech/tk543/tk545/technologies_tech_note09186
> a>
> 00800a3a25.shtml
>
> http://www.cisco.com/univercd/cc/td/doc/product/software/ios120/12cgcr/q
> os_c/qcpart4/qcpolts.htm
>
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:
> nobody@groupstudy.com] On Behalf Of
> darth router
> Sent: Monday, July 02, 2007 9:23 PM
> To: ccielab@groupstudy.com
> Subject: Shape/Police Peak
>
> Ok guys,
>
> I know this has probably been beaten to death by now,
> but I have been
> twisting my brain around it for days. Reading and
> reading, and I think I
> am finally starting to grasp the concept.
>
>
> When you police peak, you can only take tokens out of
> one bucket per
> interval, correct? If CIR is not enough, you jump to the
> Tp bucket,
> which creates the exceed action, righto?
>
> With shape peak, you can transmit from both buckets per
> interval?
>
>
> DR
>
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
> <http://www.groupstudy.com/list/CCIELab.html>
>
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
> <http://www.groupstudy.com/list/CCIELab.html>
>
>
>
>
>
> --
> Narbik Kocharians
> CCIE# 12410 (R&S, SP, Security)
> CCSI# 30832
This archive was generated by hypermail 2.1.4 : Sat Aug 18 2007 - 08:17:39 ART