Discussion in 'Tomato Firmware' started by kripz, Dec 19, 2008.
If i set some class to use max 90% and there is no other traffic, will it use 100%?
I don't think so. For inbound there is no min/max. I set "Highest" to 100%. I figure that will ensure it gets 100% if necessary, but everything else will get up to 100% if "Highest" isn't using it.
To me, it seems like anything else is caps.
I had no luck getting inbound limiting to work for me.
You can see the stats of the ingress qdisc with:
tc -s qdisc show dev vlan1
If it hasn't dropped anything then it isn't working. I think it is just the way the qos script makes the class. If you specifically set the ingress qdisc with something like:
tc filter add dev vlan1 parent ffff: protocol ip prio 5 u32 match ip src 0.0.0.0/0 police rate 6000kbit burst 15k drop flowid 1:1
then it does start to work and limit your inbound. However on the WRT54GL it seemed to make little difference on the incoming quality regardless whether or not it was set. If I was on a voip call and someone were to browse google maps it would totally trash the connection. I had identical ingress settings on a smoothwall box and I never had that problem, voip was always clear.
How do you identify your VOIP connection? I know some people who use MagicJack thought it was just a SIP connection. They said QoS didn't do anything for them because they experienced the same thing you've described. They didn't realize the actual call was carried on another connection.
And then, there were others (like me) who realized it was on another connection, used the Layer 7 RTP choice, and we thought QoS didn't do anything because we experienced the same thing you've described.
But, using a dest IP address (or the fixed, temporary rtp-2 filter), QoS kicked in.
This is also confused by the fact that different VOIP services work differently. I've tried two other softphone applications. They don't seem to use SIP or RTP. Wireshark identifies it as just UDP (or IP).
All I know is, once I saw the actual telephone call marked by QoS (in the Tomato QoS->"details"), the inbound quality was significantly improved when I saturated by connection using a www.speedtest.net test. The difference is significant compared to either disabling QoS, or changing the QoS classification rule so that it doesn't catch (and mark) that connection.
Maybe it's just me and my router. It's hard to believe that it wouldn't be at least partially noticeable by others. When someone says they don't notice it doing *anything*, I have to wonder if the rule is really catching, and if it's catching the right traffic.
No I am talking about inbound here. Outbound is fine. All voip is nothing but UDP traffic. So I just give UDP the highest priority. But browsing google maps never caused any problems with voip traffic when I was using my smoothwall box. But now with this WRT54GL setup and using the same ingress settings (the command above), I had alot of problems. Not sure why, I don't think the ingress queue is related to hardware or anything but it could be.