Re: Traffic Shaping with WFQ

Cool, I've created a single queue with no mask and pointed my rule at it so
that should take care of that.  I don't have the time to test to verify that
it actually is doing fair queuing so I'm just going to assume it's good
unless I hear otherwise.

Thanks for all your help Michael.

Just for feedback to whomever is out there the 'magic shaper wizard' wasn't
so great in my opinion.  I liked the idea of it but what I expected it to do
is what I'm doing here (a simple single fair queued rate limit of traffic
through the device).  But instead of doing that it created a pretty complex
bunch of rules which may be useful for complex scenarios.  I suppose adding
in each a single rule, pipe and queue wasn't too hard but having a 'basic
shaping' option in the wizard where I could have just put in my WAN upload
and download speed would have been handy...



On Tue, Jun 16, 2009 at 1:41 PM, Michael Sierchio <[email protected]>wrote:

> Dan Dill wrote:
>
> [Dan - I subscribe to the list, Cc: not needed.]
>
> > *"If you want all machines to share evenly a single link, you should use
> > instead: *
> >
> > *    ipfw add queue 1 ip from any to 10.1.2.0/24
> >     ipfw queue 1 config weight 5 pipe 2 mask dst-ip 0x000000ff
> >     ipfw pipe 2 config bw 300Kbit/s
> > "*
> >
> > That's what I need I think.  I would only have a single queue but I want
> to
> > ensure that bandwidth is shared equally among different connections
> within
> > that queue (and not allowing one connection to use the entire link at the
> > expense of other connections).  And a single queue seems to accomplish
> that
> > (via WFQ within the queue) whereas direct to the pipe does not according
> to
> > you.
> >
> > Is that right?
>
> Almost - by using a mask specifier, you are creating dynamic queues.  A
> new queue will be instantiated for each matching flow, with the same
> parameters as the original queue.  Yes, this is a way of ensuring that
> each flow will fairly share the bandwidth with the other flows in the
> same queue.  You could still have different queues on the same pipe
> with different weights.
>
> The FreeBSD man page for ipfw is useful, if too long.
>
>        http://tinyurl.com/mojeqy
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>