From: Volkov, Dmitry (Toronto - BCE) (dmitry_volkov@xxxxxxxxx)
Date: Thu Aug 29 2002 - 12:40:25 GMT-3
Comments inline
>
>
> Comments inline...
>
> "Volkov, Dmitry (Toronto - BCE)" wrote:
> >
> > Carlos,
> >
> > Can You point us this place "where it says BE can be larger
> than AR * Tc" ?
> >
> > Do You mean this from
> http://www.cisco.com/warp/customer/125/21.shtml ?:
> > Excess Burst Size (Be)
> > The number of non-committed bits (outside of CIR) that are
> still accepted by
> > the Frame Relay switch but are marked as eligible to be
> discarded (DE).
> >
> > The token bucket is a 'virtual' buffer. It contains a
> number of tokens,
> > enabling you to send a limited amount of data per time
> interval. The token
> > bucket is filled with Bc bits per Tc. The maximum size of
> the bucket is Bc +
> > Be. If the Be is very big and, if at T0 the bucket is
> filled with Bc + Be
> > tokens, you can send Bc + Be bits at the access rate. This
> is not limited by
> > Tc but by the time it takes to send the Be. This is a
> function of the access
> > rate.
> >
> > "If the Be is very big .." - I guess it is not about Be,
> but should be read
> > "If the BURST is very big.."
>
> NO, its ok as it is. If Be is very big, meaning that you configure
> it in the shapping as a large number.
>
> >
> > Now about this "This is not limited by Tc but by the time
> it takes to send
> > the Be. This is a function of the access rate."
> > If burst is bigger than Be+Bc, bits above Be+Bc will be
> sent as soon as
> > token bucket has enough tokens - the don't have to be sent
> within Tc - Token
> > bucket mechanism regulates when to send them
>
> As long as we have tokens available, it will transmit at AR.
> The token bucket is be+bc in size, it's loaded with bc bits
> every Tc and drained with as many bits you have to send.
>
> You do not have data to send, you can acquire as many as Be+Bc.
> You have plenty of data to send ? Then you transmit at AR untill
> you run out of tokens. As you get tokens at Bc/Tc rate, after
> you empty your token bucket, you'll continue to transmit at Bc/Tc
Nothing will allow to send number of bits above 192000 (if port speed 192K)
in ANY interval < 1 sec
Cisco DEFINITION of Be - bits above Bc during Tc. If You send Bc+Be, and
afterr that send Bc each Tc,
You don't have any tokens - wail untill 9th Tc, after that You allowed to
send again max Tc+Be, If You have more data to send, wait again if in 11th
or 12th Tc nothing was sent - You accure enogh tokens and can send mopre
than Bc, etc.
You don't have to send all bust within Tc - but you have to obey token
bucket rules.
CIR 64 K, speed 192K
Be = 192-64/8=16000 within first Tc, after that if you send each Tc 8K - You
don't have enogh tokens to send anyting more, untill 9th Tc.
> >
> > As Donny Mateo showed : each Tc interval, token bucket
> refilled with Bc
> >
> > Srinivas Vegesna IP Quality of service: page 188 :
> >
> > Burst size (Bc) is the amount of data added to the token
> bucket of size
> > (Bc+Be) at each measuring interval, Tc. Tc is defined as
> Bc/CIR. Excess
> > burst size Be is amount of excess burst of data allowed to
> be sent during
> > the first interval when the token bucket is full.
> > When a new packet arrives, it is queued into the output
> queue and scheduled
> > for transmission based on queue scheduler used, such as WFQ
> or FIFO. A
> > packet scheduled for transmission by the scheduler is
> transmitted only if
> > enough tokens are available in the bucket equivalent to the
> scheduled
> > packet. After a conforming packet is transmitted, an
> equivalent amount of
> > tokens are removed from the bucket.
> > If not all Bc bytes are sent in a Tc interval, You can
> transmit the unused
> > bytes in the subsequent interval along with the new credit
> for Bc bytes.
> > Hence, in a Tc interval during which there is less than Bc
> traffic, the
> > credit can increase to an upper bound of Bc + Be for the
> next subsequent
> > interval.
> > If serious load sets in and the token bucket is full, you
> can send Be + Bc
> > bytes in the first interval and Bc bytes in each subsequent
> interval until
> > congestion conditions ease. As you can see, you can
> throttle to the CIR
> > equation at times of congestion.
> >
> > There is very good diagram on page 189.
>
> This is what I'm also saying. This does not prevent you from having
> Be larger than what you can transmit in one Tc.
It doesn't prevent - You may put Be 1000000000 if You have speed rate 192 K
and CIR 96 K, but it will not work.
Dmitry
>
> >
> > Dmitry
> >
> > > -----Original Message-----
> > > From: Carlos G Mendioroz [mailto:tron@huapi.ba.ar]
> > > Sent: Thursday, August 29, 2002 5:49 AM
> > > To: Colin Barber
> > > Cc: Jim Brown; ccielab@groupstudy.com
> > > Subject: Re: I need FRTS help or review - TIME TO VOTE
> > >
> > >
> > > People,
> > > I'm forwarding an old message on the theme. BTW, I'm with
> option 1, as
> > > it is
> > > clearly stated (and demostrated) below. As MADMAN says,
> when in doubt,
> > > test it.
> > >
> > > Documentation in cisco is confusing to say the least, and it
> > > seems that
> > > parts of
> > > them have been reproduced "as is" many times. There is at
> least one
> > > place
> > > (also listed below) where it says BE can be larger than AR * Tc.
> > >
> > > HTH.
> > >
> > > >>>>>>
> > > Subject:
> > > Re: FRTS want to nail it down !!
> > > Date:
> > > Tue, 16 Jul 2002 17:53:19 -0300
> > > From:
> > > Carlos G Mendioroz <tron@huapi.ba.ar>
> > > To:
> > > "Paglia, John (USPC.PCT.Hopewell)"
> <JPaglia@NA2.US.ML.com>
> > > References:
> > > 1
> > >
> > >
> > > John,
> > > take a look at http://www.cisco.com/warp/customer/125/21.shtml
> > > There it's a litle more clear that the Be does not have to enter
> > > in one Tc.
> > >
> > > I really don't have much else to say, and I'm starting to
> reiterate
> > > myself, which is no good. Please do the experiment of setting the
> > > params as I said (it will take you 5 to 20 minutes) and then
> > > a lot of thinking :-)
> > >
> > > Regards,
> > > -tron
> > >
> > >
> > > "Paglia, John (USPC.PCT.Hopewell)" wrote:
> > > >
> > > > OK, now I'm getting confused on this point, because when I
> > > read Cisco
> > > > Documentation, I was led to believe that the Be is ONLY
> > > sent during the
> > > > first time cycle. Thus, for example....
> > > >
> > > > 125 ms
> > > > max= 64k
> > > > CIR= 48k
> > > > mincir= 24k
> > > > I was thinking that if the above is true, that...
> > > >
> > > > Bc= 6k (48k/8 (1sec*.125))
> > > > Be= 16k (max-above formula)
> > > >
> > > > 6k (Bc) * 8 # of time cycles) + 16k (remainder sent in
> > > first time cycle) =
> > > > 64K
> > > >
> > > > Please assist,
> > > > Pags
> > > >
> > > > > -----Original Message-----
> > > > > From: Carlos G Mendioroz [SMTP:tron@huapi.ba.ar]
> > > > > Sent: Tuesday, July 16, 2002 3:34 PM
> > > > > To: Guoqi Cui
> > > > > Cc: David Luu; Ahmed Mamoor Amimi; Bruce Williams; Lab
> > > Candidate;
> > > > > ccielab@groupstudy.com
> > > > > Subject: Re: FRTS want to nail it down !!
> > > > >
> > > > > Guoqi Cui wrote:
> > > > > >
> > > > > > As I understand CIR and BE are in different unit.
> > > > >
> > > > > Yes indeed.
> > > > >
> > > > > > We can not say set BE=32000 bps. BE should be in
> > > > > > unit of bits. As for the example, 32000 is the extra
> > > > > > link speed to handle extra traffic beyond CIR. The
> > > > > > formula to calculate BE is: (AR-CIR)*Tc. Tc is one
> > > > > > slot time interval. You can set BE higher than
> > > > > > (AR-CIR)*Tc, what is the use of it? Sinece AR is the
> > > > > > maximum rate the link can provide, after garantee CIR
> > > > > > only (AR-CIR) left.
> > > > >
> > > > > There is only AR - CIR left but in each Tc!
> > > > > You can use your "excess burst" in many Tc's.
> > > > >
> > > > > As I said in the example, even though you have only 4000
> > > > > bits left to transmit above CIR, you can actually transmit
> > > > > one full minute at AR given that you have enough Be.
> > > > >
> > > > > Please take the time to make the experiment.
> > > > > Or read some book (e.g. Tannenbaum's computer metworks)
> > > > > about the dual leaky bucket algorithm. That's what it is.
> > > > >
> > > > > Regards,
> > > > >
> > > > > >
> > > > > > As for David, you set BE as 32000bps, this is
> > > > > > different from BE's definition, you give a rate, BE
> > > > > > should be in bits.
> > > > > >
> > > > > > --- Carlos G Mendioroz <tron@huapi.ba.ar> wrote:
> > > > > > > David,
> > > > > > > I think this is not quite right, or I am
> > > > > > > misunderstanding you.
> > > > > > >
> > > > > > > To be safe, it is ok to set Be to a value >> AR/Tc.
> > > > > > >
> > > > > > > It is my understanding that Be is how much data you
> > > > > > > can transmit over
> > > > > > > your CIR,
> > > > > > > as you "eat" burst tokens. But you can eat them in
> > > > > > > more than one Tc.
> > > > > > >
> > > > > > > Here an example everyone can run to see the
> > > > > > > difference:
> > > > > > >
> > > > > > > CIR = 32 kbps
> > > > > > > Bc = 4000 b
> > > > > > > Tc = .125 s (automatically computed)
> > > > > > > Be = 1920000 b
> > > > > > > Ar = 64 kbps (Access rate, i.e. clockrate)
> > > > > > >
> > > > > > > This will transmit at 64kbps for one minute and then
> > > > > > > drop to 32k.
> > > > > > > Try it with ttcp, send 60 x 8192 bytes, it will take
> > > > > > > some 75 seconds...
> > > > > > >
> > > > > > > http://www.cisco.com/warp/customer/125/21.shtml has
> > > > > > > more info.
> > > > > > >
> > > > > > > Regards,
> > > > > > >
> > > > > > >
> > > > > > > David Luu wrote:
> > > > > > > >
> > > > > > > > like i said in a previous post...
> > > > > > > > "Be" sends excess data in the FIRST interval if
> > > > > > > there are enough tokens, it
> > > > > > > > does not send it in the other intervals
> > > > > > > >
> > > > > > > > take the bottom example...
> > > > > > > > port speed = 64000bps
> > > > > > > > cir = 32000bps
> > > > > > > > mincir = 16000bps
> > > > > > > > bc = 4000bps
> > > > > > > > be = 32000bps
> > > > > > > >
> > > > > > > > first we will start off with "Bc", to get "Bc" we
> > > > > > > divide 1 second (1000ms)
> > > > > > > > by the time interval (which is 125ms in this
> > > > > > > case), which is 1000ms divided
> > > > > > > > by 125ms = 8...this gives us 8 intervals, so a
> > > > > > > 32000bps cir divided by 8
> > > > > > > > gives us a "Bc" of 4000bps
> > > > > > > >
> > > > > > > > again, "Be" sends excess data in the FIRST
> > > > > > > interval if there are enough
> > > > > > > > tokens, it does not send it in the other
> > > > > > > intervals...to send data at the
> > > > > > > > maximum rate the port can handle, we take the port
> > > > > > > speed of 64000bps minus
> > > > > > > > the cir of 32000bps...this leaves us with 32000bps
> > > > > > > to send in the first
> > > > > > > > interval...
> > > > > > > >
> > > > > > > > if we were to have a "Be" of 4000bps, the max rate
> > > > > > > we can send is cir + be
> > > > > > > > (32000+4000), which equals 36000bps, remember that
> > > > > > > "Be" is only sent out in
> > > > > > > > the FIRST interval if there are available tokens
> > > > > > > >
> > > > > > > > with a Tc of 125ms, it gives us 8 intervals
> > > > > > > > with a Tc of 10ms, it gives us 100 intervals
> > > > > > > >
> > > > > > > > once again i have to stress this, "Be" is only
> > > > > > > sent out in the FIRST
> > > > > > > > interval if tokens are available...with a Tc of
> > > > > > > 125ms or 10ms, "Be" will
> > > > > > > > still only be sent out ONE time every second
> > > > > > > (1000ms)
> > > > > > > >
> > > > > > > > hope this clears up any confusion you may have
> > > > > > > with "Be" and "Bc"
> > > > > > > >
> > > > > > > > At 01:33 AM 7/16/2002 +0500, Ahmed Mamoor Amimi
> > > > > > > wrote:
> > > > > > > > >aaaahhhh !! atleast ur with me... heheheeh...
> > > > > > > this is what i have beening
> > > > > > > > >saying but not getting a very difinite answer
> > > > > > > from the group why they have
> > > > > > > > >used 32000 as be in power session. if bc is
> > > > > > > multipy by 0.125 then be should
> > > > > > > > >also be multiply by 0.125 ... why 1 sec
> > > > > > > > >
> > > > > > > > >-Mamoor
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >----- Original Message -----
> > > > > > > > >From: Guoqi Cui <guoqicui@yahoo.com>
> > > > > > > > >To: Bruce Williams
> > > > > > > <bruce@williamsnetworking.com>; Lab Candidate
> > > > > > > > ><labccie@yahoo.com>; Ahmed Mamoor Amimi
> > > > > > > <mamoor@ieee.org>
> > > > > > > > >Cc: <ccielab@groupstudy.com>
> > > > > > > > >Sent: Monday, July 15, 2002 9:49 PM
> > > > > > > > >Subject: RE: FRTS want to nail it down !!
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > > The FRTS parameters are quite confusing.
> > > > > > > > > > CIR, MINCIR is in unit of bps
> > > > > > > > > > BC, BE is in unit of bits
> > > > > > > > > >
> > > > > > > > > > If BE is set as: port_speed minus-cir
> > > > > > > > > > it will hava a unit of bps,
> > > > > > > > > >
> > > > > > > > > > Do Tc is assumed to be 1 sec?
> > > > > > > > > >
> > > > > > > > > > For this case( Tc=0.125 sec),
> > > > > > > > > >
> > > > > > > > > > BE should be set as 32000/8=4000.
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > --- Bruce Williams
> > > > > > > <bruce@williamsnetworking.com>
> > > > > > > > > > wrote:
> > > > > > > > > > > If the available bandwidth is 32000, then
> > > > > > > dont we
> > > > > > > > > > > still need to multiply it
> > > > > > > > > > > by .0125 to get the bits per Tc?
> > > > > > > > > > >
> > > > > > > > > > > -----Original Message-----
> > > > > > > > > > > From: nobody@groupstudy.com
> > > > > > > > > > > [mailto:nobody@groupstudy.com]On Behalf Of
> > > > > > > > > > > Lab Candidate
> > > > > > > > > > > Sent: Sunday, July 14, 2002 1:03 AM
> > > > > > > > > > > To: Ahmed Mamoor Amimi
> > > > > > > > > > > Cc: ccielab@groupstudy.com
> > > > > > > > > > > Subject: Re: FRTS want to nail it down !!
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > Mamoor,
> > > > > > > > > > >
> > > > > > > > > > > This is obvious, since you said the port
> > > > > > > speed is
> > > > > > > > > > > 64000 and cir is 32000,
> > > > > > > > > > > if you want to utilized the most of your
> > > > > > > available
> > > > > > > > > > > bandwidth for burst
> > > > > > > > > > > purpose, the be would be equal to port_speed
> > > > > > > minus
> > > > > > > > > > > cir, hence 32000 bps.
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > I was reading the power-session 2002, here
> > > > > > > on page
> > > > > > > > > > > 66 they have given the
> > > > > > > > > > > values like :
> > > > > > > > > > >
> > > > > > > > > > > Port Speed: 64000 Bps
> > > > > > > > > > > CIR : 32000 Bps (Average Rate of Traffic
> > > > > > > Without
> > > > > > > > > > > Congestion)
> > > > > > > > > > > Mincir: 16000 Bps (Average Rate of Traffic
> > > > > > > With
> > > > > > > > > > > Congestion)
> > > > > > > > > > > Bc: 4000 Bps (Amount of Data Sent Per
> > > > > > > Interval)
> > > > > > > > > > > Be: 32000 Bps (Amount of Excess Allowed Once
> > > > > > > Credit
> > > > > > > > > > > Has Built Up)
> > > > > > > > > > >
> > > > > > > > > > > interface Serial0
> > > > > > > > > > > ip address 10.10.10.1 255.255.255.0
> > > > > > > > > > > encapsulation frame-relay
> > > > > > > > > > > frame-relay traffic-shaping
> > > > > > > > > > > frame-relay class ccie
> > > > > > > > > > >
> > > > > > > > > > > map-class frame-relay ccie
> > > > > > > > > > > frame-relay adaptive-shaping becn
> > > > > > > > > > > frame-relay cir 32000
> > > > > > > > > > > frame-relay mincir 16000
> > > > > > > > > > > frame-relay bc 4000
> > > > > > > > > > > frame-relay be 32000
> > > > > > > > > > >
> > > > > > > > > > > Can someone explain why Be = 32000 , why not
> > > > > > > Be =
> > > > > > > > > > > 32000 * 0.125 = 4000
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > >
> > > > > >
This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:48:42 GMT-3