Re: Shaping and calculating Be value for a 30 second burst

This is fine AFAIK, and inline with the first example.
But don't take my word for it, it is very easy to set up a lab and
test this! You can see that, if everything is ok, you can send @ 128
for 20 seconds and then it just drops to 64k. Do it!

(I've done that myself, take a look at
http://news.groupstudy.com/archives/ccielab/200207/msg01417.html)

-Carlos

Muzammil Malick @ 26/04/2010 7:07 -0300 dixit:
> So from my last point I think I have confused you
>
> If I had the following config
>
> map-class frame-relay shape
> frame-relay cir 64000
> frame-relay bc 8000
> frame-relay be 1920000
>
> given a Tc of 125ms and an AR of 128000 is this ok to allow for bursting
> to AR for 30seconds or have I just typed a load of rubbish?
>
> By the way, Ladeegeek I am SO SORRY for hijacking your post but this has
> been bugging me for ages.
>
> On 26 April 2010 10:51, Carlos G Mendioroz <tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar>> wrote:
>
> Again, inline:
>
> Muzammil Malick @ 26/04/2010 5:19 -0300 dixit:
> > Carlos
> >
> > This makes a lot of sense now. I was not thinking about Be as a
> quantity
> > but as a value you can burst each
> > time interval, but this was obviously wrong.
> > When I started to think about the Be as a value that is saved and then
> > burst I came up with the same result as your previous email:
> > /
> > /
> > /Given that you have Bc=8000, with just 8000 bits more per Tc you can
> > reach AR. To keep your Tx at AR for 240 Tcs you need 240 x 8000,
> > or 1.920.000./
> > /
> > /
> > However based on this I am assuming that the answer in the
> original post
> > is wrong:
> > /
> > /
> > /AR - 1.544 mb
> > Cir - 768k
> > Tc - 125 ms
> > burst for 30 sec up to line rate/
> > /
> > /
> > /The answer key for this task states that the Be is calculated by
> taking
> > the line rate and multiplying it by 30 sec?
> >
> > 1544000bps*30s = 46320000bits/
> > /
> > /
> > The correct answer would be Bc=96000. To Tx to AR per Tc we need
> another
> > 97000bits.
> > To Tx to AR for 30secs(8x30=240) we would have 97000x240=23280000bits
>
> This sounds fishy, and reveals that someone does not have some
> experience with circuits. 1.544Mb is the raw speed of a T1, but
> 1 every 193 bits is framing, so T1 actual usable speed is 1.536Mb,
> which is 2 * 768. Just an aside, but worth mentioning...
>
> >
> > The final point I am unsure about is this:
> >
> > If we configured a Be value as above, 23280000bits does this mean that
> > you can accumulate 23280000bits in a Be bucket
> > and once this has all be burst and no more credit has accumulated then
> > there are no more Be bits available to burst?
>
> Yep, that's the idea. You need to fall below CIR to regain credit, and
> you regain the same way yo expend it. If you fall to 0bps, you would
> earn 96000 bits per Tc, if you go down to 384k, then only 48000, and
> so on.
> If you are of a math type, it's the area between your tx and the CIR
> line. Ouch, it's an integral :)
>
> >
> > Thus 23280000 is the amount that can be accumulated and burst for a 30
> > second period, whether you send this in a continual sustained
> burst for
> > 30 seconds or whether you sent this over a period of 30 seconds in a 2
> > hour period purely depends on when you have accumulated credit?
>
> Hmm, I don't understand this, but if you were to tx at 1000bps and had
> full credit of 23280000 bits, that would let you go for some 100 seconds
> before you have to go back to CIR.
>
> -Carlos
> >
> >
> >
> >
> >
> > On 26 April 2010 01:40, Carlos G Mendioroz <tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar>
> > <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>> wrote:
> >
> > Please see inline:
> >
> > Muzammil Malick @ 25/04/2010 19:53 -0300 dixit:
> > > Hi
> > >
> > > I, ok so I was one of the below until a few days ago:
> > >
> > > /if you ask legendary network engineers (60+ y.o!) they'd
> say "Bc
> > is CIR
> > > per TC
> > > and Be is AIR-CIR per Tc". Actually their training was
> simple and
> > > effective...hence again you can say both Bc and Be are referring
> > to the
> > > number
> > > of bits and not rates. Technically it's true to say my Bc or
> my Be
> > is 32000
> > > bits but it's not true when people say hey the Bc is now
> 32Kbps./
> > >
> > > but after reading emails on GS I have come to understand
> that I do not
> > > really understand this aswell as I should.
> > >
> > > Carlos, I answered a question on traffic shaping the other
> day in
> > which
> > > it stated "allow burst up to port speed if credit has been
> > accumulated"
> > > You replied to the same thread saying that because Be was not Tc
> > related
> > > the question should state how long Be applies for.
> > > In the question in this thread the question states that
> bursting is
> > > possible for 30 seconds.
> > >
> > > My question is how does stating how long we can burst up to
> line rate
> > > affect the Be?
> >
> > The math of how much you are allowed to transmit is done
> synchronically,
> > i.e., at regular intervals that we call Tc.
> > Say Tc is 1 sec. Bc = 8 means you can send 8 bits (1 byte) per
> > second (Tc).
> > Now, this is fine if you transmit at the same time all the
> time, but
> > what if you tx in a bursty way ? Well, you may be allowed to
> save some
> > tx allowance by transmitting at a lower pace (say, not
> transmitting at
> > all). How much can you save ? Be.
> >
> > If you have a Be of 1000, can you tx at 1Kps ? May be, may be not.
> > Because you have another limit which is your access rate.
> >
> > >
> > > e.g if AIR=128k CIR=64k Bc=8000
> > >
> > > if we can Be to port speed for 1 second the Be will be 8000,
> if we can
> > > Be to port speed for 30 seconds the Be will be 8000.
> >
> > Hmm, first thing I would calculate is Tc. This is CIR/Bc, i.e.
> 125ms.
> > Being you AR 128k, that means that every Tc you can transmit,
> at most,
> > 16000 bits. And in 1 second there are 8 Tcs, in 30 seconds
> there would
> > be 240 Tcs.
> > Given that you have Bc=8000, with just 8000 bits more per Tc
> you can
> > reach AR. To keep your Tx at AR for 240 Tcs you need 240 x 8000,
> > or 1.920.000.
> >
> > > Doesn't this all just depend on how much credit is accumulated?
> >
> > Yes, and that is Be exactly. Be is not how much you can
> overspeed in
> > one Tc...
> >
> > If this sounds complicated, just go stare the Stallings book cover
> > until you get the idea of how this works. :)
> >
> >
> http://www.pearsonhighered.com/educator/academic/product/0,,0139737448,00%2ben-USS_01DBC.html
> >
> > -Carlos
> >
> > >
> > > Thanks in advance
> > >
> > > On 25 April 2010 22:56, Carlos G Mendioroz <tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar>
> > <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>
> > > <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>
> <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>>> wrote:
> > >
> > > I would second the move to understand it.
> > > Bc is Tc related. Tc is your clock, Bc is how much
> credit you get
> > > at every tick to transmit.
> > > Be is not Tc related, so can be much larger than Bc, or
> even 0.
> > > And it is how much credit you can save as long as you spend
> > (transmit)
> > > less than your Bc bits. You can not spend all of it at
> once, since
> > > no matter how much credit you have, AR is a hard limit
> on the
> > speed
> > > of the link.
> > >
> > > Now Bc/Tc is a base rate that is accounted (CIR). If you
> want
> > to rev up,
> > > your only need credit for the part over CIR up to AR. Hence
> > the AR-CIR
> > > count.
> > >
> > > -Carlos
> > >
> > > Ladee Geek @ 25/04/2010 16:10 -0300 dixit:
> > > > I have a question about determining the Be for generic
> traffic
> > > shaping.
> > > > I've looked through the QoS exam guide, QoS
> configuration guide,
> > > the doc cd
> > > > and haven't seen enough to be able to figure this one out.
> > > >
> > > > AR - 1.544 mb
> > > > Cir - 768k
> > > > Tc - 125 ms
> > > > burst for 30 sec up to line rate
> > > >
> > > > I understood Be to be AR - Cir.
> > > >
> > > > The answer key for this task states that the Be is
> calculated by
> > > taking
> > > > the line rate and multiplying it by 30 sec?
> > > >
> > > > 1544000bps*30s = 46320000bits
> > > >
> > > > There's just no way to put that many bits on the wire
> in 1 sec.
> > > Or is the
> > > > Be the number of tokens to be replenished in 1 sec (
> same as
> > 5790000
> > > > bits/Tc)
> > > >
> > > > And while I am at it, if you've found a good reference
> > please pass
> > > it on.
> > > >
> > >
> > > --
> > > Carlos G Mendioroz <tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar>
> > <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>
> <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>
> > <mailto:tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>>>
> > > LW7 EQI Argentina
> > >
> > >
> > > Blogs and organic groups at http://www.ccie.net
> > >
> > >
> >
> _______________________________________________________________________
> > > Subscription information may be found at:
> > > http://www.groupstudy.com/list/CCIELab.html
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> >
> > --
> > Carlos G Mendioroz <tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar> <mailto:tron_at_huapi.ba.ar
> <mailto:tron_at_huapi.ba.ar>>>
> > LW7 EQI Argentina
> >
> >
>
> --
> Carlos G Mendioroz <tron_at_huapi.ba.ar <mailto:tron_at_huapi.ba.ar>>
> LW7 EQI Argentina
>
>

-- 
Carlos G Mendioroz  <tron_at_huapi.ba.ar>  LW7 EQI  Argentina
Blogs and organic groups at http://www.ccie.net