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

Re: new WRR sched, updated TBF and "real" ingres implementation

On Fri, 21 Jul 2000 devik@cdi.cz wrote:

> the WRR can't be simulated by prio+TBFs. With WRR you can
> divide whole BW into several classes (bands) where each
> class has several percent from BW.
> But when one (or more) of classes is unused the remaining
> BW is distributed among all other classes which
> have something to send (it is distributed again in relative
> percentage of quantums).

This can be done with CBQ with a single parent and multiple children
emanating from the same parent branch. Give them varying weight and the
same priority. I havent looked at your code, but CBQ in fact uses DRR
which is a better algorithm.