Hello gentleman, i wrote a few days ago but did not receive any answers. I simplified the problem and even wrote a quick example which demonstrates the innacurate rate precision: tc qdisc del dev eth0 root >/dev/null 2>&1 tc qdisc add dev eth0 root handle 1: htb tc class add dev eth0 parent 1: classid 1:101 htb rate 2048kbit tc qdisc add dev eth0 parent 1:101 handle 101: sfq perturb 10 tc filter add dev eth0 protocol ip parent 1: prio 5 handle 101 fw flowid 1:101 iptables -A PREROUTING -t mangle --src <ip_address> -j MARK --set-mark 101 The example above should shape the speed to 2mbit (2048kbit) but unfortunately the real speed is only 1920kbit.. Tthe only solution I have found is to add the missing rate. Linux 2.4.20 vanilla. Any comments? P.Krumins _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Am Saturday 01 November 2003 21:32 schrieb Peteris Krumins:> i wrote a few days ago but did not receive any answers.I did not receive/read this mail.> The example above should shape the speed to 2mbit (2048kbit) > but unfortunately the real speed is only 1920kbit..How exactly do you measure the ''real speed''? The difference sounds much like overhead of some sort that you''re not accounting properly. For example, when using scp to copy files over the line, it would be normal to have lower transfer rates than the actual line speed, because there''s some kind of handshake/encryption overhead with every package sent. Regards, Andreas Klauer _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Monday, November 3, 2003, 2:37:26 AM, you wrote: AK> Am Saturday 01 November 2003 21:32 schrieb Peteris Krumins:>> i wrote a few days ago but did not receive any answers.AK> I did not receive/read this mail. That one was about limits and data precision.>> The example above should shape the speed to 2mbit (2048kbit) >> but unfortunately the real speed is only 1920kbit..AK> How exactly do you measure the ''real speed''? The difference sounds much like AK> overhead of some sort that you''re not accounting properly. For example, when AK> using scp to copy files over the line, it would be normal to have lower AK> transfer rates than the actual line speed, because there''s some kind of AK> handshake/encryption overhead with every package sent. Via ftp. If I perform the tests using tbf attached to a class then the rate is precise. For example, if I measure rate of 256kbit using sfq and then tbf, sfq is always wrong at (actual_rate/rate_set) 240kbit/256kbit = 0,9375 = 6,25% 1920kbit/2048kbit = 0,9375 = 6,25% I performed several tests at different data rates and the innacuracy of 6,25 percent always remain. P.Krumins _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/