Does traffic shaping increase bandwidth usage?
I configured traffic shaping to prioritize VOIP traffic. However, I realized that my monthly bandwidth usage increased a lot. I suspect it could be related to the traffic shaper. Is it something possible? Is there a way to avoid that? The problem is that my monthly bandwidth is limited by my ISP, and I must pay for exceeding traffic.
I run 1 network with 20 some + users . I set up the shaper then added rules and queues for each of the 10 top hogs. I had it set so each user could get max 5 Mb down and max 256 Kb up for 1 minute then cut them back too 1 Mb and 64 kb .
The internet usage almost doubled. After 3 months I killed the shaper and usage dropped back to normal. I'm talking close too 200 Gb with out this type of shaping and close to 400 Gb with the shaper as I had it configured.
Pf 1.2.3 not 2.0
The traffic accountant in both cases is located at ISP's side (or at least, outside of pfSense). Data amount may be (and actually it was) increased because shaping means that some packets must be delayed or dropped:
– The former reduces data amount if you're using full bandwidth at full time, which is unlikely.
-- The later increases data amount if the dropped packets are resent, which is very probable.
However 100% increment (i.e. 50% drop) is simply too much. There must be something wrong either in the shaper configuration, in the shaper code, or in the traffic accountant itself.
I remember seeing drops in the thousands for most of the users.
My attempt with the shaper was to stop the hogs from taking all the bandwidth downloading torrents.
Witch worked to the point that regular usage was close to normal, but the invoice from the ISP was outrageous as i said close too double the usage from the 2 previous months.
That PC I was using cooked it's board and i replaced it with a box running pf 2.0 rc1 back in march.
It's working for what I need and I'll leave it be until rc4 or a full release of 2.0 comes out.
I do run the shaper on 1.2.3 on another system to prioritize game traffic and it seems to do the job pretty well. I do get some glitches from time to time with port 80 traffic hogging.
and for some reason queue Games up shows drops ?
Perhaps there is some major change in 2.0 RC2 traffic shaper that made it less predictable than RC1. It needs time to stabilize.
This is really interesting. I think that dropped packets are the way that TCP is supposed to limit the bandwidth used, but maybe it requires some "tuning" of the queue lengths? Maybe if there are hundreds of connections for each user which I understand can be the case with P2P, then when limiting the bandwidth each connection keeps getting dropped packets to slow them down, but because there are so many separate TCP flows, each one is hardly getting anything through, although of course the ISP does see all the traffic. Maybe necessary to limit simultaneous connections per user? or per queue that is?
I also thought, maybe with the limits, everyone's browsing experience was so much better, they use it more! But you said usage was close to normal…
Very informative thank you. It makes sense that An adverse effect of shaping is increased data on the ISP's side. Because delayed data gets re transmitted because it thinks it's "in error". Where as is has a long response or timeout.
It's good information for others to concider. You may be limiting you internet usage going out but hitting your WAN, it's increased because of the drop retransmit.