[ previous ] [ next ] [ threads ]
 From:  Adam Nellemann <adam at nellemann dot nu>
 To:  John Voigt <megeek at wyndhamhills dot com>
 Cc:  m0n0wall at lists dot m0n0 dot ch
 Subject:  Re: [m0n0wall] QoS for VoIP
 Date:  Fri, 25 Jun 2004 22:32:14 +0200
John Voigt wrote:
> The queue weights prioritize the traffic through the pipe.  If you make the
> pipe smaller than your available bandwidth then you'll never reach your
> available bandwidth unless you have some other pipes that use it.
> John Voigt, President

Yes, but what about when a queue on some other device (router, modem, 
whatever) becomes saturated (which can happen if the shaper pipe is 
larger than the actual bandwidth)? Won't such a situation (in some 
circumstances, such as when it is the incomming traffic that is 
saturated) prevent m0n0wall from shaping properly, as it will then be 
that other, saturated, device that is "responsible" for throtling the 
bandwidth (such as by dropping packets), and since that device doesn't 
"honor" the m0n0wall shaper rule/pipe/queue setup, shaping will be 
arbitrary (ie. like without the m0n0wall shaper enabled) until the 
device is again un-saturated, and the m0n0wall shaper pipes and queues 
are once again the only "bottleneck".

I might still be missing something, but unless there is more to it 
than the part about being able to use the whole bandwidth, I would 
personally still choose to forgo those 1-5% of my bandwidth, if it 
means the shaper will work better (especially since the above problem 
typically occur when shaping is most important, as in when the line is 
near saturation). Also, I get the impression that one always tend to 
loose a little bandwidth when using a traffic shaper (or was that just 
the latency getting somewhat higher?)