Hi there, I have done a lot of testing and I have a few questions about CBQ. You can split the traffic in diffent classes and it works great. But to do this, I have to add as the first class a bounded class with rate = available link-bandwidth. I have tried different rates but CBQ isn't very accurate. I added the testing results at the bottom of the posting (or go to http://users.belgacom.net/staf/) As you can see there are some very strange results (notice the change in throughput when I changed the rate 1kbps around 71475kbps and 135556kbps). I also noticed that the accuracy of splitting the traffic in different classes is very good unless you want to give a class < 25% of the bandwidth. I'm wondering if someone encountered the same problem and/or if it is a problem with tc or the kernel itself. -- Staf More QOS info : http://users.belgacom.net/staf/ RESULTS : RATE_TOT RESULT 1 kbps 1.05 KB/s 2 kbps 2.10 KB/s 3 kbps 3.04 KB/s 4 kbps 4.17 KB/s 5 kbps 5.02 KB/s 10 kbps 9.91 KB/s 15 kbps 14.7 KB/s 20 kbps 18.8 KB/s 30 kbps 29.2 KB/s 40 kbps 36.5 KB/s 50 kbps 49.2 KB/s 60 kbps 49.4 KB/s 65 kbps 49.3 KB/s 67 kbps 49.3 KB/s 68 kbps 49.3 KB/s 69 kbps 49.3 KB/s 71475 bps 51.1 KB/s 71476 bps 69.7 KB/s 70 kbps 69.9 KB/s 80 kbps 69.9 KB/s 90 kbps 70.3 KB/s 100 kbps 75.4 KB/s 130 kbps 75.7 KB/s 135556 bps 78.7 KB/s 135557 bps 139 KB/s 140 kbps 140 KB/s 150 kbps 140 KB/s 200 kbps 140 KB/s 300 kbps 140 KB/s 500 kbps 140 KB/s 1000 kbps 140 KB/s 1310719bps 146 KB/s 1310720bps 940 KB/s 2000 kbps 945 KB/s
