[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Classification of AF DP's to GRED DP

Hi all,

If i remember it right (i traced it down a long time ago), the last two bits (00->VQ1, 01->VQ2 etc) of the tcindex value are used for the selection of the virtual queue. So the sample script shifts/masks the bits so the correct DSCP is mapped to the correct AF PHB.

By the way, a more flexibel way of selecting the VQ could be quite handy here (hint :)).


jamal wrote:

On Thu, 22 Feb 2001, Panagiotis Stathopoulos wrote:

> Hello all,
> I am trying to understand the mechanism of classifying different AF Drop
> precedence from the same class (i.e AF11, AF12 etc...) to different Drop
> Precedences at the GRED scheduler. I have studied and run (with out
> sucess I may admit) some relevant examples and I have n't still
> understand the mechanism this mapping works. Any help, or tested script
> would be appreciated,
> Thanks in advance,

Hi Panagiotis,

If you are refering to the gred AF examples in the scripts,
they are slightly tricky. Even I have to sit and scratch my head a few
times when i havent seen them in a while. Werner and I cooked them at some
point late at night under caffeine influnce.
Basically the initial tcindex is based on the DSCP value of the packet
but is manipulated by dsmark a few times before finally reaching GRED
which maps it to the correct GRED virtual queue. You might have to trace
it, since no notes on it exist. This could be a good use of Werner's new
user space toy.
I apologize. Note, however, we just wanted to get some sample examples and
not to preach it as the bible. As Larry Wall says (i think he said this),
"there's more than one way to do it". The sample scripts were tested
extensively for correctness.


Steven Van den Berghe             
Workgroup Broadband-Networks
Department Information Technology 
Ghent University - Belgium
Phone:  +32 (0)9 267 35 86 | Fax  :  +32 (0)9 267 35 99
The software said Windows95 or better, so I got a Linux!