Chris Hoy Poy wrote:
> no, its *extremely* effective at smoothing the traffic when you have it set
> near the right bandwidth. Dont forget that the traffic shaper can only have
> limited effect on incoming traffic anyway, as those packets are sent from the
> upstream provider.
> I had a few problems with ADSL routers that had low buffers, and normal
> traffic to them would flood them so much they'd disconnect from the net (so
> the net connection would see-saw under load). The traffic shaper effectively
> stopped this, and under load the traffic would be nicely smoothed (and still
> responsive, so the acks/nacks were getting thru in a timely fashion etc).
> On Wednesday 25 October 2006 09:00, Mark Ryan wrote:
>> Chris Hoy Poy wrote:
>>> hello again ;)
>>> You've got pipes and queues.
>>> my understanding is this:
>>> Queues represent possible slices of a pipe. The shaper will try to allow
>>> each queue its *minimum* % bandwidth of a pipe, but the queue is allowed
>>> to use all of the pipe if its available.
>>> But pipes represent a hard limit, to my knowledge, and wont borrow from
>>> other pipes. You need to do that allocation at the queue level.
>>> Rules then allocate actual traffic to each queue.
>>> so looking at your question again : if FTP is in the low priority queue
>>> and everything else is in the high priority queue, then FTP will use the
>>> maximum pipe speed, but will drop back down to its minimum when something
>>> else happens.
>>> Note the one "semi" issue I have seen is where you allocate a pipe to be
>>> less then the bandwidth you have available - it seems to use a bucket
>>> system to allocate the pipe, so if you allocate a really small pipe (like
>>> 512kbit) and you actually have 2mbit of bandwidth there, the pipe will
>>> flood the bandwidth for 1/4 of the time available and the actual usage
>>> will be very jagged. That may not apply, but I've done it a few times
>>> without thinking and found the behaviour (though it makes sense when I
>>> think about it) at the time was a bit bizarre. There might be some way to
>>> change the interval that the traffic shaper uses? It seems to be set at
>>> like 5 or 10 seconds. Having smaller intervals would make the flow a bit
>>> smoother, I assume.
>>> does that make sense? ;)
>>> On Tuesday 24 October 2006 09:44, Mark Ryan wrote:
>>>> I've been evaluating whether the m0n0wall traffic shaper will work for
>>>> me and I have to say wow! It seems very nice.
>>>> I do have 1 question:
>>>> Does the traffic shaper use borrowing? Such as if i have 2 upload ques,
>>>> ftp assigned to low priority que and everything else assigned to high
>>>> prio que. Will the ftp consume the maximum available pipe speed until
>>>> something in the high prio ques needs it, then i would assume the ftp
>>>> traffic would slow down until the bw is available again?
>>>> Sorry for the cryptic question there.
>>>> To unsubscribe, e-mail: m0n0wall dash unsubscribe at lists dot m0n0 dot ch
>>>> For additional commands, e-mail: m0n0wall dash help at lists dot m0n0 dot ch
>> Yes, makes sense. That jagged use of bw bothers me though. Does that
>> happen if you set the pipe close to the bw available?
>> Thanks for the explanation!
Excellent. I hope to be using it on a 10mbit/1mbit Cable connection. I
am just trying to justify to myself that spending $215 on the WRAP box
is a good deal. :) I have an IPCOP box that does really well right now
but im tired of the noise and heat and the failing fans and harddrives.
In 4 years ive lost 3 fans and 1 hardrive. Plus the power consumption
of a machine sitting idle all the time while routing isnt very attractive.
Once I convince myself to buy it, im sure I will enjoy it.