GroupStudy.com GroupStudy.com - A virtual community of network engineers
 Home  BookStore  StudyNotes  Links  Archives  StudyRooms  HelpWanted  Discounts  Login
Re: Traffic shaping: burst size, excess question posted 01/19/2003
[Chronological Index] [Thread Index] [Top] [Date Prev][Date Next] [Thread Prev][Thread Next]


Just to throw something into the mix. The time interval is not always 1/8 of
Cir. Cisco recommends an interval of 1/10 of CIR for voice. t/c = .1

----- Original Message -----
From: <wolfec@xxxxxxxxxxxxxx>
To: "Joe Chang" <changjoe@xxxxxxxxxxxxx>
Cc: <ccielab@xxxxxxxxxxxxxx>
Sent: Saturday, January 18, 2003 6:32 PM
Subject: Re: Traffic shaping: burst size, excess question


> Joe,
>
> From everything I've read Bc is 1/8 of CIR (not minCIR). So in you're
example
> of 64kbps CIR and 32kbps minCIR the Bc value would be 8kbps.
>
> It is the Be value that is hard to find any documentation on. One example
I had
> given to me was to subtract the CIR from AR (physical port speed to arrive
at
> Be). So if you have a port speed/AR of 128kbps and the CIR is 64kbps then
Be is
> equal to 0. Cisco sent me something stating that if I had a port speed/AR
of
> 128kbps and a CIR of 64kbps then Be should equal 72kbps for the first Tc.
>
> So if I am following this thread correctly then if I had a port speed/AR
of
> 256kbps and a CIR of 64kbps the Be should equal 200kbps for the first Tc.
>
> -Chuck
>
>
>
> Joe Chang wrote:
>
> > Thanks Scott. I agree that there's no cut and dried way to work this
thing
> > in the real world. However what I'm getting at is that I can't find any
> > guidance in the literature on how to choose values for Bc and Be. Under
> > which conditions would you want a larger or smaller Bc value, or is it
> > entirely arbitrary? How do you take the physical port transmission limit
to
> > derive the value of Be, or is that even necessary?
> >
> > ----- Original Message -----
> > From: "Scott Morris" <swm@xxxxxxxxxx>
> > To: "'Joe Chang'" <changjoe@xxxxxxxxxxxxx>; "'Aidan Marks'"
> > <amarks@xxxxxxxxx>
> > Cc: <ccielab@xxxxxxxxxxxxxx>
> > Sent: Saturday, January 18, 2003 6:08 PM
> > Subject: RE: Traffic shaping: burst size, excess question
> >
> > > Much of that question depends on your interpretation of what CIR is,
and
> > > what you are engineering to.  All of these are really network design
> > > questions.
> > >
> > > Do you engineer assuming you'll get full port speed and want to never
go
> > > below your contracted rate?   (in which case MinCIR = Bc/Tc and CIR =
Bc
> > > + Be (aka port rate))
> > >
> > > Do you engineer assuming you normally transfer at your contracted
rate,
> > > but still want to handle BECN's (perhaps your total CIR of all PVCs
> > > exceeds actual port speed).  In that case CIR = Bc/Tc and AR = Bc + Be
> > > and MinCIR = 1/2 * CIR
> > >
> > > In the lab, you either make something up, or work with the parameters
> > > given.  In real life, it will depend on many circumstances of what you
> > > are trying to achieve.
> > >
> > > Hope that helps some!
> > >
> > > Scott
> > >
> > > -----Original Message-----
> > > From: nobody@xxxxxxxxxxxxxx [mailto:nobody@xxxxxxxxxxxxxx] On Behalf
Of
> > > Joe Chang
> > > Sent: Saturday, January 18, 2003 3:39 PM
> > > To: Aidan Marks
> > > Cc: ccielab@xxxxxxxxxxxxxx
> > > Subject: Re: Traffic shaping: burst size, excess question
> > >
> > >
> > > Thanks for the response Aidan. My response inline:
> > >
> > > > >
> > > > >My question is that if you are given the following values: CIR
> > > > >64kbps,
> > > minCIR
> > > > >32 kbps, how would you begin to calculate a Bc and Be values that
> > > > >would
> > > allow
> > > > >take full advantage of the CIR?
> > > >
> > >
> > > > is 64kb the port speed? if so, the cir = port speed, so you can send
> > > > 8000 bits per 125 msec.  Since you are sending at port speed as bc,
> > > > the Be is zero (i.e you can not burst past the port speed).
> > >
> > > I should have been more specific. Say 64 kbps is the port speed and 32
> > > kbps is the contracted average rate.  Be could be greater than 0 in
this
> > > case.
> > >
> > > > if the port speed was actually 128K and you did this:
> > > >
> > > > map-class frame-relay 128KB
> > > >   frame-relay cir 64000
> > > >   frame-relay bc 8000
> > > >   frame-relay be 64000
> > > >   frame-relay mincir 32000
> > > >   frame-relay adaptive-shaping becn
> > > >
> > > > you would be sending out like this:
> > > >
> > > > 72000(Bc+Be), 8000, 8000, 8000 ....
> > > >
> > > > Be is the amount of data you can send in the *first* interval.  Once
> > > > the credit is used up, you can only send at Bc until an idle
interval
> > > > or a partially used interval when you build up credit again.  The
> > > > maximum
> > > credit
> > > > is Be.
> > >
> > > In your case of cir 64 kbps and mincir 32 kbps, on the router which
> > > value would be used to calculate the interval? Is it :
> > >
> > > CIR = Bc/Tc ,  or minCIR= Bc/Tc
> > >
> > > If we use CIR then Tc would equal 125 msec. The transmit rate in the
> > > first interval of your example would be 72 kbps  / 125 msec = 576
kbps,
> > > which is above the port speed of 128 K. Would this be possible?
> > >
> > > If we use minCIR then Tc would equal 250 msec. The transmit rate of
the
> > > burst excess interval would be 72 kbps  / 250 msec = 288 kbps , also
> > > above the port speed of 128 K.
> > >
> > > Best,
> > >   Joe
> .
.
__________________________________________________________________
To unsubscribe from the CCIELAB list, send a message to
majordomo@xxxxxxxxxxxxxx with the body containing:
unsubscribe ccielab