From: Chris Lewis (chrlewiscsco@gmail.com)
Date: Sun May 21 2006 - 11:18:36 ART
Victor,
The key point in this discussion is to focus on the ability of the router to
utilize the Be value. In either shape average or shape peak, Be can only be
used if enough credit has been built up. So, if over a long period of time,
the offered load is at or above CIR, there is no difference, as no credit
can build up.
If the offered load runs at less than the CIR for an extended period of
time, credit has been built up and there is the potential for different
behavior. Shape average will only allow you to use this credit up once per
second, whereas shape peak will allow you to use the credit at every time
interval if you have the credit available.
Both wil shape to the target rate over a long period. Be is not intended to
let you transmit at a rate above CIR for an extended period and does not
support that.
I don't know of practical applications for the shape peak behavior.
Chris
Chris
On 5/20/06, Petr Lapukhov <petrsoft@gmail.com> wrote:
>
> Victor,
>
> AFAIK, that "shape peak" stuff is just a convenient way to write down your
> SP's policy.
>
> Actually, it shapes your traffic up to "targer rate" with
> CIR1=CIR*(1+Be/Bc)
>
> *constantly*. That is, it's almost the same as setting "shape average"
> with
> CIR1, Bc1=Bc+Be, Be1=0.
>
> The only convenience is that you may clearly write down what rate is
> "guaranteed" by your SP (average), and what is the maximum possible
> "oversubscription" rate (target rate).
>
> That is, you send up to "peak" rate any time, but your SP only guarantees
> delivery of "average" packet rate. It's up to you to live with that policy
> :)
>
> Also, you may use adaptive shaping to react to congestion.
>
> HTH
> Petr
>
> 2006/5/21, Victor Cappuccio <cvictor@protokolgroup.com>:
> >
> > Hola a todos.
> >
> >
> >
> > Guys
> >
> >
> >
> > I know this had been discussed before, and that QoS Gurus in this list
> > like
> > Chris Lewis had responded the difference between shape average and shape
> > peak.
> >
> >
> >
> > But what if we look at 1 hour of shape average and 1 hour shape peak
> > (Sorry
> > I do not have a proper lab to test this using MRTG) for a particular
> flow,
> > What will be in the difference when we send traffic using peak and then
> > sending the traffic but now using average?.
> >
> >
> >
> > I know that BE+BC are only sent at the First Time Interval using shape
> > average and with the use of shape peak we can send BC+BE every TC (off
> > course if we have the tokens in the bucket). But my question is,. .. -
> > Should I send more traffic using Peak Parameter in the Shape command
> > then??
> >
> >
> >
> > Imagine only one flow of HTTP Traffic (steady and constant) going out
> the
> > interface and then we apply shape average for 1 hour, and then apply
> shape
> > peak for other hour, how could this be showed in the MRTG Graff?
> >
> >
> >
> > How could the wording (besides using the average or peak) could be
> > elaborated for both scenarios?
> >
> >
> >
> > Thanks in Advance and sorry for my bad English
> >
> > Victor.
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Thu Jun 01 2006 - 06:33:22 ART