From: Colin Barber (Colin.Barber@xxxxxxxxxxxxxx)
Date: Wed Aug 28 2002 - 15:53:03 GMT-3
This 9th Tc is also a bit unclear.
The token bucket size is BC+BE. Tokens are added at the rate of BC each Tc.
For example if BC=8000 and BE=20000 then bucket size is 28000. If you have
1Mb of data being switched to the frame interface, 28000 bits are sent on
the first Tc and the rest sit in the FIFO queue (or whatever queue is
configured). The number of tokens is now 0. At the next Tc 8000 tokens are
placed into the bucket and therefore 8000 bits are transmitted. If Tc=1/8
sec then after 1 second 84000 bits have been transmitted.
On the next Tc (which is the 9th) only 8000 tokens are placed into the
bucket and therefore only BC is sent. You never fill up the bucket to
transmit above BC until the FIFO queue has been emptied at which point the
bucket will start to fill up to more than just BC. This is why the CIR is
the average transfer rate. You may be able to send 84k in one second but the
average over time will come down and down to almost the CIR which is 64k.
Colin
-----Original Message-----
From: Volkov, Dmitry (Toronto - BCE) [mailto:dmitry_volkov@ca.ml.com]
Sent: 28 August 2002 19:07
To: 'Jim Brown'; 'Colin Barber'; 'Omer Ansari'
Cc: 'ccielab@groupstudy.com'
Subject: RE: I need FRTS help or review
Jim,
You are right traffic will be delayed and sent within 9th Tc interval
(assuming Tc=1/8 sec).
But Be is still Be - it is number of bits, and Be+Bc =< Port Speed
(clocking) * Tc.
We have to put actual number in config.
40000 bits will not be transmitted during 1st Tc - only 12000 bits.
Other bits will be delayed until 9th Tc.
Be - number of bits will be transmitted within 1st Tc ==> be = 12000
-8000(Bc) = 4000 bits
and this number 4000 we have to put in config as Be. The other bits will be
delayed / dropped, whatever.
I'm not 100% sure about it, but I got this logic from CCO.
Colin right - the documentation on this subj is very unclear.
This subj is like religion for me - believe or not believe :)
Dmitry
> -----Original Message-----
> From: Jim Brown [mailto:Jim.Brown@caselogic.com]
> Sent: Wednesday, August 28, 2002 11:35 AM
> To: 'Colin Barber'; 'Omer Ansari'; 'ccielab@groupstudy.com'
> Subject: RE: I need FRTS help or review
>
>
> You do not necessarily need 320Kps available during the first
> interval. The
> packets are software queued before they are placed onto the
> TxRing. This is
> the whole idea behind shaping.
>
> They are not placed on the wire at exactly the same speed
> designated per TC
> interval.
>
> List, if I'm way off base please correct me and help me see the light.
>
> -----Original Message-----
> From: Colin Barber [mailto:Colin.Barber@telewest.co.uk]
> Sent: Wednesday, August 28, 2002 9:09 AM
> To: 'Omer Ansari'; 'ccielab@groupstudy.com'
> Subject: RE: I need FRTS help or review
>
>
> What's the access speed of the interface? If it's 96k then this is
> incorrect. In the first time slot BC+BE=40000. For that
> amount of data to be
> transferred in the first timeslot the access speed will need to be
> 40000*8=320kbps.
>
> Colin
> -----Original Message-----
> From: Omer Ansari [mailto:omer@ansari.com]
> Sent: 28 August 2002 10:56
> To: ccielab@groupstudy.com
> Cc: steven.j.nelson@bt.com; Jim.Brown@caselogic.com;
> kip.palmer@verizon.net
> Subject: RE: I need FRTS help or review
>
>
> Guys,
>
> so bottom line is Jim and Steve have correctly done this right?
> It seems good to me, and I too want to stick this once and for all.
>
> omer
>
> On Mon, 26 Aug 2002 steven.j.nelson@bt.com wrote:
>
> > All
> >
> > Jim is correct in this one, his figures pan out as follows
> >
> > CIR 96000
> > MINCIR 64000
> > BE 32000
> > BC 8000
> > TC 0.125Ms
> >
> > So in 8 time slots (1 Second) he will transmit :-
> >
> > 0.125Ms 40000 (BC+BE)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> > 0.125Ms 8000 (BC)
> >
> > Which is equivalent to 96K per second.
> >
> > And when no tokens are available then the MIN CIR will be
> met by the 8000
> BC
> > * 8 = 64000
> >
> > Thanks to Jim for this one.
> >
> > Steve
> >
> >
> > -----Original Message-----
> > From: Jim Brown [mailto:Jim.Brown@caselogic.com]
> > Sent: 26 August 2002 15:24
> > To: 'kpalmer'
> > Cc: 'ccielab@groupstudy.com'
> > Subject: RE: I need FRTS help or review
> >
> >
> > After your e-mails I rethought my stance on FRTS. I did a
> little more
> > research and I believe my original configuration at the
> bottom of the post
> > is the correct answer from a lab or testing context for a
> 96Kbps port and
> > 64Kbs contracted CIR.
> >
> > map-class frame-relay TestShape
> > frame-relay cir 64000
> > frame-relay be 32000
> > frame-relay bc 8000
> >
> > I'm basing this on a single new piece of information I
> turned-up. Check
> the
> > Networkers 2002 CCIE Power Session, in their FRTS example,
> they configure
> > the parameters exactly as I have described below.
> >
> > I still stand by my original assessment of Cisco's CIR set to the
> providers
> > CIR and Cisco's BE set to the difference between providers
> CIR and port
> > speed.
> >
> > I'm posting this back to the list to hopefully open up
> discussion again.
> >
> > -----Original Message-----
> > From: kpalmer [mailto:kip.palmer@verizon.net]
> > Sent: Sunday, August 25, 2002 8:27 PM
> > To: 'Jim Brown'
> > Subject: RE: I need FRTS help or review
> >
> >
> > Line speed | Access Rate | Port Speed
> > =======================================
> > What you bought from the Provider. Per DLCI.
> >
> >
> > Average Rate | configured CIR (not mincir)
> > =======================================
> > When Shaping 128 to 64, it's 64k, with Bc ='s the Average
> Rate of remote
> > 64, /8.
> >
> > -----Original Message-----
> > From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]
> On Behalf Of
> > Jim Brown
> > Sent: Sunday, August 25, 2002 1:04 PM
> > To: ccielab@groupstudy.com
> > Subject: I need FRTS help or review
> >
> >
> > Everything I have read about FRTS doesn't seem to clear up
> the use of
> > BE, BC, CIR, and MINCIR. I have been unable to locate a
> solid resource
> > explaining the concept with any finality.
> >
> > I've read most of the relevant Usenet postings on Deja, watched the
> > threads on groupstudy, scoured CCO, and examined the QOS v1.0 course
> > material.
> >
> > I will throw out my assumptions and let list members either
> verify or
> > shoot holes on my take of FRTS.
> >
> >
> > A few definitions up front:
> >
> > AR is the Access Rate or Port Speed of the connection to
> the frame relay
> > cloud. This is the maximum number of bits that can be
> transmitted to the
> > cloud.
> >
> > CIR is the Committed Information Rate. This is the maximum number of
> > bits the provider promises to transmit. Anything above the
> CIR and below
> > the access rate will have the DE bit marked and is eligible for
> > discard/drop during times of congestion.
> >
> > Lets take a hypothetical circuit for instance, a port speed
> of 96Kbps
> > and a CIR of 64Kbps.
> >
> > The way I read the documentation, in a Cisco configuration
> CIR should be
> > set to the actual provider CIR or 64000. The BE or burst
> excess should
> > be set to the difference between the access rate and the
> CIR. I think BE
> > should be set to 32000, the difference between 96 and 64.
> >
> > Here is a brief sample config:
> >
> > map-class frame-relay TestShape
> > frame-relay cir 64000
> > frame-relay be 32000
> >
> >
> > The map-class could then be applied to the frame map or the
> interface. I
> > was previously under the impression you would set the Cisco
> CIR to the
> > port speed and the minCIR to the provider contracted CIR. I
> don't think
> > this is really the case?
> >
> > Here is an example:
> >
> > map-class frame-relay TestShape
> > frame-relay cir 96000
> > frame-relay mincir 64000
> >
> > Comments or suggestions? Is this wrong, why or why not?
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:48:40 GMT-3