Re: OT: Shaping and extended burst

I was referring to the Be = Bc or such simple lab rules.

If you have the source of the one you are mentioning, it would be nice
to know the basis of such rule. I can not figure out why the round trip
time would have any relation to this, but may be in a good design
the numbers happen to match ?

I would expect it to be more sensible to the ratio between AR and CIR.
As I said, this is in the end compensating jitter. Comonsense applies,
so to say. Also, may be the app is bursty to begin with! Too many ifs.

-Carlos

Chris Proctor @ 05/03/2011 13:03 -0300 dixit:
> The best I have been able to find is be = 2(traffic per round trip time)
>
> So, if we expect a maximum round trip time of 100ms and 100kb to be sent
> during that 100ms... the be would be 200kb. Is that more or less your
> understanding?
>
> On 3/5/2011 10:31 AM, Carlos G Mendioroz wrote:
>> Well, if Be is too small the app will suffer.
>> If it is too large, in the extreme, you don't shape at all.
>> But the "right" one is dependent on so many things that we have
>> some rules to do something that is not too bad.
>>
>> -Carlos
>>
>> Chris Proctor @ 05/03/2011 12:06 -0300 dixit:
>>> Nod. This is on topic for me.
>>>
>>> I believe Aaron's interpretation is correct. I can understand why
>>> the burst is needed, but I have never been able to figure out
>>> "correct" burst size and what really happens when the burst size is
>>> too large or small.
>>>
>>> Chris
>>>
>>> On 3/5/2011 8:29 AM, Aaron Riemer wrote:
>>>> Thanks Carlos,
>>>>
>>>> I just wanted to know if my understanding was correct more than
>>>> anything.
>>>> When things don't add up I like to question rather than just accept an
>>>> answer without knowing why. This I think helps to solidify my
>>>> understanding.
>>>>
>>>> Thanks for your help again mate.
>>>>
>>>> Cheers,
>>>>
>>>> -Aaron
>>>>
>>>> -----Original Message-----
>>>> From: nobody_at_groupstudy.com [mailto:nobody_at_groupstudy.com] On Behalf Of
>>>> Carlos G Mendioroz
>>>> Sent: Saturday, 5 March 2011 7:57 PM
>>>> To: Aaron Riemer
>>>> Cc: ccielab_at_groupstudy.com
>>>> Subject: Re: OT: Shaping and extended burst
>>>>
>>>> Aaron,
>>>> > What I am confused about is the fact that more than the CIR rate can
>>>> > actually be sent over the initial lets say 1 second ...
>>>>
>>>> Now think about it, it's only the *first* second. When did it happen in
>>>> an actual implementation ? :)
>>>>
>>>> It's called *burst* for a reason. And doing the analysis from the math
>>>> is not going to help you understand it, IMHO.
>>>>
>>>> It's there to compensate the jitter in the packets, because your
>>>> line rate is/may be way faster than your CIR, and your "small window
>>>> perception" (aka Tc) of the rate may not be accurate.
>>>>
>>>> As far as I see it, this is not OT at all.
>>>>
>>>> -Carlos
>>>>
>>>> Aaron Riemer @ 05/03/2011 04:34 -0300 dixit:
>>>>> Hi Guys,
>>>>>
>>>>>
>>>>>
>>>>> I am just reading up on the token bucket principles with shaping in
>>>>> particular with the use of Bc and Be (extended burst capability).
>>>>>
>>>>>
>>>>>
>>>>> I have read that when Be is configured the token bucket maximum
>>>>> size will
>>>>> equal Bc + Be. Therefore initially (Let's say the first Time Period
>>>>> Tc) Bc
>>>> +
>>>>> Be packets will be taken from the bucket.
>>>>>
>>>>>
>>>>>
>>>>> Does this mean that when Be is configured you can initially send
>>>>> more than
>>>>> the actual shaped rate?
>>>>>
>>>>>
>>>>>
>>>>> Take this example:
>>>>>
>>>>>
>>>>>
>>>>> I enable traffic shaping and set the shaped rate to 64kbps over a
>>>>> 128kbps
>>>>> serial link. I stick with the default Tc and therefore calculate Bc
>>>>> to be
>>>>> 8000bits (64000 * .125). I enable Be to be 8000bits as well.
>>>>>
>>>>>
>>>>>
>>>>> Going by the token bucket logic the shaper will take (Bc+Be) from the
>>>> bucket
>>>>> (the bucket is full to start with). This means 16000 bits are taken
>>>>> effectively allowing the burst to 128kbps for the first Tc of .125ms.
>>>>>
>>>>>
>>>>>
>>>>> What I am confused about is the fact that more than the CIR rate can
>>>>> actually be sent over the initial lets say 1 second of transfer (8
>>>>> * Tc).
>>>> I
>>>>> calculate this to be 72k bits for the first second of transfer and
>>>>> then
>>>> 64k
>>>>> bits for every subsequent second assuming full load over one second.
>>>>>
>>>>>
>>>>>
>>>>> Is this correct? I am just trying to get my head around it and
>>>>> understand
>>>>> this completely.
>>>>>
>>>>>
>>>>>
>>>>> Any comments welcome as always.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Thanks guys,
>>>>>
>>>>>
>>>>>
>>>>> -Aaron
>>>>>
>>>>>
>>>>> 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>  LW7 EQI  Argentina
Blogs and organic groups at http://www.ccie.net