[altq 187] Re: CBQ question

[Salvatore Iacono <iacono@athena.polito.it>]

Prashant Chandra wrote:

> Kenjiro Cho wrote:
> >
> > > Shouldn't the TCP stream on class4 also be able to use all the
> > > banwidth allocated to class3 (because they are under the same
> > > parent class class1)?  I expected to see about 5 Mpbs for TCP in
> > > this case too --- if hierarchical link sharing was working right.
> >
> > The link-sharing mechanism of CBQ is different from that of HFSC.
> > CBQ does not have a clear rule on how the excess bandwidth is
> > distributed.  You cannot expect excess bandwidth distribution
> > is proportional to assigned bandwidth.
> >
> > -Kenjiro
>
> There is excess bandwidth and there is unused bandwidth.  I
> think one of the goals of hierarchical link sharing in CBQ,
> is to distribute bandwidth unused by a class to its children
> in the hierarchy.  How much of the unused bandwidth each child
> gets may not be clearly defined.
>
> Still, I find it odd that class3's banwidth is not allocated
> to anybody else.  There is about 4 Mbps of unused bandwidth
> that neither of the flows use.  Is the CBQ scheduler non
> work-conserving?
>
>   -prashant
> --

Yes, the CBQ queue is non work-conserving. For more details, see
"cbq-howto.txt" in the "docs" folder of altq distribution.
-salvatore