From: Petr Lapukhov (petrsoft@gmail.com)
Date: Thu Jun 15 2006 - 04:53:00 ART
Pierre,
Could you please monitor flow weights with
"show queue serial x/y"
while pinging? :)
Petr
2006/6/15, Pierre-Alex <paguanel@hotmail.com>:
>
> Petr,
>
> I have repeated the testing with WFQ in the class-default.
>
>
> Class1---- > ping precedence 5 ---> 50% of bandwidth
> Class2---> ping precedence 3 ----> 20% of bandwidth
> Class default ---> WFQ ---> ping of other precedence
>
> The results are the following:
>
> If packets have a precedence of zero, they get no bandwidth.
>
> However I increased the precedence to 4, the packets get about 1 K of
> bandwidth.
>
> I would have expected the default class to get at least 5% of bandwidth
> remaining: that is 6.4 K
>
> But this is not the case !!!!
>
> Pierre-Alex
>
>
> ----- Original Message -----
> *From:* Petr Lapukhov <petrsoft@gmail.com>
> *To:* Pierre-Alex <paguanel@hotmail.com>
> *Cc:* Chris Lewis <chrlewiscsco@gmail.com> ; ccielab@groupstudy.com
> *Sent:* Wednesday, June 14, 2006 6:06 PM
> *Subject:* Re: Findings: Behaviour of Default Class with Class Based
> Shaping
>
> Pierre, you asked an interesting question.
>
> Actually, I still wonder how exactly class-default is assigned it's
> weight...
>
> It's been told in docs, that if you don't put a "banwidth" of "fair-queue"
> in
> class-default, it is treated as "best-effort".
>
> That sounds to me "class-default" served after every other "weighted"
> queue in that case. And it looks like it works that way, in your examples.
>
> Well, if you do put "bandwidth" or "fair-queue", class-default
> is being integrated into WFQ algorithm, and is served accordingly.
>
> I would try to investigate that matter further :)
>
> Petr
This archive was generated by hypermail 2.1.4 : Sat Jul 01 2006 - 07:57:32 ART