GroupStudy.com GroupStudy.com - A virtual community of network engineers
 Home  BookStore  StudyNotes  Links  Archives  StudyRooms  HelpWanted  Discounts  Login
RE: Shape average and shape peak posted 10/25/2006
[Chronological Index] [Thread Index] [Top] [Date Prev][Date Next] [Thread Prev][Thread Next]


I am not sure where you are getting that Tc is a constant value.  I
would agree that for a given CIR and Bc, the Tc is constant, but I do
not agree if you are saying that Tc is always a defined value.  If it
were, VoIP would not work.  For instance, what if a SP provided you a
10MB link but you only paid for 3MB.  You would have to force your
traffic not go over 3MB.  Also, if utilizing VoIP, you would want to
make sure that your Tc was around 10ms.  Therefore, you would need to
shape.

Using CIR = Bc/Tc (Avg), you would need to solve for Bc since you know
your values for the CIR and the value you want for Tc.  

So, 3MB = Bc/10ms  ==> Bc = 30000

Now, if you are using Shape Peak to use the full 10MB link (Which you
would never do in the voice arena), you would have:

7MB = 3MB (1 + Bc/Be)  ==> 7Mb = 3Mb (1 + Be\(3Mb * Tc))

Let our Tc = 10 ms as in the shape average...

Solve for Be ==> 39990

However, since Be is randomized, you would always want to set it to 0 in
real life scenarios.  If you don't, you could have excess jitter.  That
is why I always use shape average.

This being beside the point, the original point of my post was to point
out the difference between using shape peak with a given bc and be and
using shape average and setting the Bc to the same value as bc + be in
shape peak.

Vince Mashburn
Voice / Data Engineer
901-263-5072
CCVP, CCNP, CCDA,Network +
Cisco IP Telephony Support Specialist
Cisco IP Telephony Operations Specialist
-----Original Message-----
From: WorkerBee [mailto:ciscobee@xxxxxxxxx] 
Sent: Wednesday, October 25, 2006 3:35 AM
To: Vincent Mashburn
Cc: alexeim@xxxxxxxxxxxxxxx; Cagri Yucel; Group study
Subject: Re: Shape average and shape peak

With either shape average or shape peak, the Tc calculation is fixed.

Shape average -  CIR = Bc / Tc
Shape peak     -  CIRpeak = CIR (1 + Be/Bc)
                                       = Bc/Tc + Be/Tc
                                       = 1/Tc (Bc + Be)    note : Tc is
constant

Use shape average if the CIR must always maintained.

Use shape peak if your network allows you to burst above CIR.

Be value range from 0 to (Access Rate - Bc).

Hence, both commands are meant for different meaning, should not
just view it from a mathematically view. It does not work that way.


On 10/25/06, Vincent Mashburn <vmashburn@xxxxxxxxx> wrote:
> It is my understanding that Be only gets transmitted only if excess
> bandwidth is available, even in shape peak.  So, if your transmitted
> traffic totally saturates the entire token bucket at a particular
> interval (Tc), then there is no room to use the Be, so the transmitted
> Be is 0 (the entire Bc is used).  However, if the Bc did not take the
> entire bucket for that interval, the Be is allowed to be used until
the
> excess bandwidth is used up.  Now for Shape average, this Be is
> calculated based on the previous intervals (Tc) information whereas
the
> shape peak allows the Be to transmitted based on the current intervals
> (Tc) information.  This is how I understand it.  Please add any
> additional info to this that you can.
> Thanks
> Vince Mashburn
> Voice / Data Engineer
> 901-263-5072
> CCVP, CCNP, CCDA,Network +
> Cisco IP Telephony Support Specialist
> Cisco IP Telephony Operations Specialist
>
> -----Original Message-----
> From: Alexei Monastyrnyi [mailto:alexeim@xxxxxxxxxxxxxxx]
> Sent: Tuesday, October 24, 2006 12:47 PM
> To: Vincent Mashburn
> Cc: Cagri Yucel; Group study
> Subject: Re: Shape average and shape peak
>
> Hi Wincent.
>
> When adding to Bc we definitely increase Tc with the same CIR... but
> under the hood... Odom in DQOS Guide says "shape average" behavior is
> the same as normal GTS behavior, i.e spending Be after "a long period
of
>
> inactivity" during which Be is being accumulated, whereas "shape peak"
> spends tokens from Be on every Tc.
>
> If we spend tokens from Be at each and every Tc and our traffic
profile
> is not lower than conform rate, i.e >=CIR all the time, where do we
get
> Be accumulated from?
>
> I have a feeling that this Be is not the same as normal
traffic-shaping
> Be... It is just a sort of extra credit we've got to be able to burst
> more at each Tc ... without increasing a Tc...
>
> Does it make any sense?
>
> A.
>
>
> Vincent Mashburn wrote:
> > This is a common question.  To answer the question, refer back to
the
> > formula: CIR = bc/tc.
> >
> > For shape peak, bc + be is sent every interval, but the formula is
> still
> > CIR = bc / tc.  For shape average, bc is sent every interval, and
the
> > formula is still CIR = bc / tc.
> >
> > So, say you have a CIR of 64000, bc = 8000 and be = 16000.
> > Using shape average:
> >
> > 64000 = 8000 / tc  ==> tc = 125ms.
> >
> > Using shape peak:
> >
> > 64000 = 8000 / tc ==> tc = 125ms
> >
> > Using shape average by setting bc = bc + be
> >
> > 64000 = 24000 / tc ==> tc = 375ms.
> >
> > So, as you can see, it is different.
> > Hope this helps.
> >
> > Vince Mashburn
> > Voice / Data Engineer
> > 901-263-5072
> > CCVP, CCNP, CCDA,Network +
> > Cisco IP Telephony Support Specialist
> > Cisco IP Telephony Operations Specialist
> >
> > -----Original Message-----
> > From: nobody@xxxxxxxxxxxxxx [mailto:nobody@xxxxxxxxxxxxxx] On Behalf
> Of
> > Cagri Yucel
> > Sent: Tuesday, October 24, 2006 4:49 AM
> > To: Group study
> > Subject: Shape average and shape peak
> >
> > I am sure I am missing a bit here but on the Cisco Doc it says
> >
> > shape average sends Bc bits in each interval
> > shape peak sends Bc+Be bits in each interval
> >
> > So what's the point of having a different command ? Isn't it the
same
> > thing
> > if I set shape average with a Bc equal to Bc+Be of shape peak ?
> >
> > Thanks
>
>
_______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html