[altq 1132] Re: Equal bandwidth distribution among hosts in asubnet using ALTQ

[Kenjiro Cho <kjc@csl.sony.co.jp>]

Nicolas Christin wrote:
> Syazwan Hisyam bin Mohamed Ridzwan wrote:
> 
> > Is there any way that I can define the parent class to use CBQ and use WFQ on 
> > the child class?
> 
> 
> 	I don't think you can do that with a single interface, but if you cascade 
> two interfaces, the first running CBQ, and the second one running WFQ, 
> you may be able to achieve what you want to do. In other words, assuming 
> you have four classes of traffic what you could have is:

[snip]

I thought diffserv would provide much simpler solutions than complex
combinations of different queueing disciplines but diffserv deployment
has been slower than I expected...

> > I've done something quite silly yesterday. Just want to ask your opinion on 
> > this. I plugged in 4 fxp (Intel eepro) network card onto one FreeBSD machine. 
> > Set bridging (using sysctl) on fxp0 & fxp1. Then set bridging using netgraph 
> > on fxp2 & fxp3 (I don't set bridging using sysctl on fxp2 & fxp3 to avoid 
> > mute when loop is detected). I use cross network cable on fxp1 & fxp2. fxp0 
> > is connected to the local network. fxp3 is connected to the outside world. 
> > Then, I use CBQ on interface fxp2 and WFQ on fxp0. It seems like no problem 
> > happened, everyone here can still access the Internet until now. But I'm 
> > curious whether WFQ on fxp0 really works or not... OR... should I think of 
> > different ways to shape the traffic???
> 
> 	
> 	If I'm not mistaken, I think this is pretty much what I just described 
> above. It should work, but you can make sure by using altqstat -i 
> interface and examining the output.

The tricky part is that, if you cascade multiple queues with the same
bandwidth links, only the first queue can be backlogged.
And a queueing discipline works only when the queue is backlogged.

-Kenjiro