RE: [LARTC] CBQ question

[Gery Kahn <geryk@sphera.com>]

It seems that configuration like
      qdisc(CBQ)
       |
     class
   |		|
 class1   class2
 bounded  bounded

doesn't work perfect.The more stable ( accurate ) configuration would be
    qdisc(cbq) 1:0
	|
    class 1:1
	|
    qdisc(cbq) 2:0
    |          |
  class 2:1  class 2:2
    |
  qdisc(tbf) 3:0
Now when class 2:1 have rate 10Mbit and class 2:2 = 70Mbit
i can have 9.05 and 72.774

> Please.
> I create simple config.
>
>     CBQ
>        |
>    class
>  | |
> class1   class2
>
> when rate of class1 is high for the rate of class2,
> finaly rate of class2 exceed the given.
> My experiments gives for example if
> class1 rate = 90Mbit and class2 rate = 10Mbit so
> actual rate of class2 can be from 11.4 to 12.5 ( and far more ).
>
> Thank you.
>
 I test for CBQ and get the same result.
This question I asked before, it's maybe weight, avpkt, cbq problem...
I change the avpkt to 500Kbit(I don't know why???) and get better accuracy,
maybe you can try this.
Another way is change your queueing disciplines(cbq) to tbf, it's really
more stable than cbq, it won't get jitter when your avpkt is change.


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/2.4Routing/