Dear all, I ve set CEIL parameter each class not more than 128Kbit But sometimes the CEIL parameter can be higher 128Kbit As far as i monitored, this CEIL rate can be 132Kbit maximum, dont know why.. Regards, Rio Martin. -- May the Fleas of a Thousand Camels infest one of your Erogenous Zones. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hello again Rio Martin, : I ve set CEIL parameter each class not more than 128Kbit But sometimes : the CEIL parameter can be higher 128Kbit As far as i monitored, this : CEIL rate can be 132Kbit maximum, dont know why.. For how long does the monitored output traffic exceed your ceiling? It is possible that you''ll exceed your ceil briefly because the allotted traffic can exceed ceiling by the amount of cburst. If you are recording an output rate above your ceil (128kbit) consistently over a long period of time (a large transfer), then I, also, do not know why. [ Stef may have more to say about this. ] Also, if you want a hard cap on your bandwidth at 128kbit, you might try setting cburst to zero. Alternatively, you can drop the ceil parameter to a setting just below 128kbit--this is my preferred technique. -Martin -- Martin A. Brown --- SecurePipe, Inc. --- mabrown@securepipe.com _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 23 July 2003 12:40, you wrote:> Hello again Rio Martin, > : I ve set CEIL parameter each class not more than 128Kbit But sometimes > : the CEIL parameter can be higher 128Kbit As far as i monitored, this > : CEIL rate can be 132Kbit maximum, dont know why.. > For how long does the monitored output traffic exceed your ceiling? It is > possible that you''ll exceed your ceil briefly because the allotted traffic > can exceed ceiling by the amount of cburst. > If you are recording an output rate above your ceil (128kbit) consistently > over a long period of time (a large transfer), then I, also, do not know > why. > [ Stef may have more to say about this. ] >This is lasts for long period.. Usually happened when hosts in that class perform download or online streaming.> Also, if you want a hard cap on your bandwidth at 128kbit, you might try > setting cburst to zero. Alternatively, you can drop the ceil parameter to > a setting just below 128kbit--this is my preferred technique.I tried to set to zero: class htb 1:3003 parent 1:4 leaf 3003: prio 0 rate 4Kbit ceil 24Kbit burst 0b cburst 0b class htb 1:1000 parent 1:2 leaf 1000: prio 0 rate 32Kbit ceil 48Kbit burst 0b cburst 0b rate 4332bps 5pps backlog 52p results: still exceed around 50 - 80 bps Regards, Rio Martin. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 23 July 2003 07:40, Martin A. Brown wrote:> Hello again Rio Martin, > > : I ve set CEIL parameter each class not more than 128Kbit But sometimes > : the CEIL parameter can be higher 128Kbit As far as i monitored, this > : CEIL rate can be 132Kbit maximum, dont know why.. > > For how long does the monitored output traffic exceed your ceiling? It is > possible that you''ll exceed your ceil briefly because the allotted traffic > can exceed ceiling by the amount of cburst. > > If you are recording an output rate above your ceil (128kbit) consistently > over a long period of time (a large transfer), then I, also, do not know > why. > > [ Stef may have more to say about this. ]132kbit - 128kbit = 4kbit and that''s not much. So I don''t think you have to worry. Is this ceil parameter on a leaf class (a class with no child classes)? Or can you post you script? Also, execute tc -s -d clas sshow dev eth0 and watch the ctokens. They should never be negative for that class. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hello, I had configure my QoS router and have two computers conected to it, one of then genreates traffic and the other recive the traffic. I would like test the HTB and CBQ qdisc. What software can I use to generate traffic? What software can I use to analyse and graphic the traffic received? How can I analyse the traffic that past trought the two interfaces in the router? Thanks in advance. Ana _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 23 July 2003 16:30, Ana Carolina Alonso wrote:> Hello, > I had configure my QoS router and have two computers conected to it, one > of then genreates traffic and the other recive the traffic. > I would like test the HTB and CBQ qdisc. What software can I use to > generate traffic? What software can I use to analyse and graphic the > traffic received? > How can I analyse the traffic that past trought the two interfaces in the > router?I have some script that 1. creates the traffic, 2. do realtime logging and 3. creates graphs. You can find some of these scripts on www.docum.org. I use ttcp to generate traffic (sometimes ab or ping) and the iptables counters to count the packets. Classification is done on dst port. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 23 July 2003 21:16, Stef Coene wrote:> On Wednesday 23 July 2003 07:40, Martin A. Brown wrote: > > [ Stef may have more to say about this. ] > 132kbit - 128kbit = 4kbit and that''s not much. So I don''t think you have > to worry. > Is this ceil parameter on a leaf class (a class with no child classes)? Or > can you post you script? > Also, execute tc -s -d clas sshow dev eth0 and watch the ctokens. They > should never be negative for that class.Yes, the ceil parameter on leaf class. I tried Martin''s way by setting Burst and Cburst value to zero. And the value of bps exceed CEIL parameter around 50 - 80 bps. Thanks. Regards, Rio Martin. -- It will be advantageous to cross the great stream ... the Dragon is on the wing in the Sky ... the Great Man rouses himself to his Work. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Thursday 24 July 2003 04:22, Rio Martin. wrote:> On Wednesday 23 July 2003 21:16, Stef Coene wrote: > > On Wednesday 23 July 2003 07:40, Martin A. Brown wrote: > > > [ Stef may have more to say about this. ] > > > > 132kbit - 128kbit = 4kbit and that''s not much. So I don''t think you have > > to worry. > > Is this ceil parameter on a leaf class (a class with no child classes)? > > Or can you post you script? > > Also, execute tc -s -d clas sshow dev eth0 and watch the ctokens. They > > should never be negative for that class. > > Yes, the ceil parameter on leaf class. > I tried Martin''s way by setting Burst and Cburst value to zero. > And the value of bps exceed CEIL parameter around 50 - 80 bps.Can you post your script and the output of tc -s -d class show dev eth0? Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Thursday 24 July 2003 23:28, you wrote:> On Thursday 24 July 2003 04:22, Rio Martin. wrote: > Can you post your script and the output of tc -s -d class show dev eth0?tc qdisc del dev eth0 root tc qdisc add dev eth0 root handle 1 htb default 0 r2q 10 tc class add dev eth0 parent 1: classid 1:10 htb rate 40Kbit ceil 40Kbit burst 1 cburst 1 tc qdisc add dev eth0 parent 1:10 handle 10 sfq perturb 10 quantum 1500 tc class add dev eth0 parent 1:10 classid 1:9100 htb rate 20Kbit ceil 20Kbit burst 1 cburst 1 tc filter add dev eth0 parent 1:0 protocol ip prio 200 handle 100 fw classid 1:9100 root@noc:/etc/sysconfig/htb# tc -s -d class show dev eth0 class htb 1:10 root rate 40Kbit ceil 40Kbit burst 0b/8 mpu 0b cburst 0b/8 mpu 0b level 7 Sent 1406352 bytes 12398 pkts (dropped 0, overlimits 0) rate 532bps 2pps lended: 0 borrowed: 0 giants: 0 tokens: -7521 ctokens: -7521 class htb 1:9100 parent 1:10 leaf 9100: prio 0 quantum 1000 rate 20Kbit ceil 20Kbit burst 0b/8 mpu 0b cburst 0b/8 mpu 0b level 0 Sent 1406352 bytes 12398 pkts (dropped 0, overlimits 0) rate 504bps 2pps lended: 12398 borrowed: 0 giants: 0 tokens: -15041 ctokens: -15041 Regards, Rio Martin. -- He who attacks the fundamentals of the American broadcasting industry attacks democracy itself. -- William S. Paley, chairman of CBS _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Tuesday 29 July 2003 07:28, Rio Martin. wrote:> On Thursday 24 July 2003 23:28, you wrote: > > On Thursday 24 July 2003 04:22, Rio Martin. wrote: > > Can you post your script and the output of tc -s -d class show dev eth0? > > tc qdisc del dev eth0 root > tc qdisc add dev eth0 root handle 1 htb default 0 r2q 10Bad idea. Class 0 is a special one and I dont''t think you want unclassified traffic go out at full speed ....> tc class add dev eth0 parent 1: classid 1:10 htb rate 40Kbit ceil 40Kbit > burst 1 cburst 1Argh. Remove the burst and cburst parameters so htb can calculate the smalles one for you. I think that''s your problem.> tc qdisc add dev eth0 parent 1:10 handle 10 sfq perturb 10 quantum 1500 > tc class add dev eth0 parent 1:10 classid 1:9100 htb rate 20Kbit ceil > 20Kbit burst 1 cburst 1 > tc filter add dev eth0 parent 1:0 protocol ip prio 200 handle 100 fw > classid 1:9100 > > root@noc:/etc/sysconfig/htb# tc -s -d class show dev eth0 > class htb 1:10 root rate 40Kbit ceil 40Kbit burst 0b/8 mpu 0b cburst 0b/8 > mpu 0b level 7 > Sent 1406352 bytes 12398 pkts (dropped 0, overlimits 0) > rate 532bps 2pps > lended: 0 borrowed: 0 giants: 0 > tokens: -7521 ctokens: -7521 > > class htb 1:9100 parent 1:10 leaf 9100: prio 0 quantum 1000 rate 20Kbit > ceil 20Kbit burst 0b/8 mpu 0b cburst 0b/8 mpu 0b level 0 > Sent 1406352 bytes 12398 pkts (dropped 0, overlimits 0) > rate 504bps 2pps > lended: 12398 borrowed: 0 giants: 0 > tokens: -15041 ctokens: -15041Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Rio Martin.
2003-Jul-29 08:21 UTC
How to deal with burstable bandwidth from ISP ? Was: Re: Over value in CEIL parameter..
On Tuesday 29 July 2003 14:45, you wrote:> > tc qdisc add dev eth0 root handle 1 htb default 0 r2q 10 > Bad idea. Class 0 is a special one and I dont''t think you want > unclassified traffic go out at full speed ....That because i dont want to shape my ssh connection & other connection from my pc to bandwidth.limiter (: I am still learning about fw mark to shape both incoming and outgoing bandwidth from my network. Please let me know how to shape incoming rate with Iptables fw mark I ve done with the outgoing bandwidth but i still dont understand how to mark incoming bandwidth from my ISP. example: /usr/sbin/iptables -t mangle -A POSTROUTING -s 192.168.1.0/24 -j MARK --set-mark 100 this is for upgoing connection from my local network. tc class add dev eth0 parent 1: classid 1:100 htb rate 40Kbit ceil 40Kbit tc qdisc add dev eth0 parent 1:100 handle 100 sfq perturb 10 quantum 1500 tc filter add dev eth0 parent 1:0 protocol ip prio 200 handle 100 fw classid 1:100> > tc class add dev eth0 parent 1: classid 1:10 htb rate 40Kbit ceil 40Kbit > > burst 1 cburst 1 > Argh. Remove the burst and cburst parameters so htb can calculate the > smalles one for you. I think that''s your problem.if i remove those parameters, it will exceed about 4 - 5 Kbit if i use those parameters, it only exceed about 40-80 bit (: About my ISP, Downstream connection is burstable. I get CIR 512Kbps burstable to 1024Kbps Please advise how to deal this situation Stef, Thanks. Regards, Rio Martin. -- To A Quick Young Fox: Why jog exquisite bulk, fond crazy vamp, Daft buxom jonquil, zephyr''s gawky vice? Guy fed by work, quiz Jove''s xanthic lamp -- Zow! Qualms by deja vu gyp fox-kin thrice. -- Lazy Dog _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Stef Coene
2003-Jul-29 08:34 UTC
Re: How to deal with burstable bandwidth from ISP ? Was: Re: Over value in CEIL parameter..
On Tuesday 29 July 2003 10:21, Rio Martin. wrote:> On Tuesday 29 July 2003 14:45, you wrote: > > > tc qdisc add dev eth0 root handle 1 htb default 0 r2q 10 > > > > Bad idea. Class 0 is a special one and I dont''t think you want > > unclassified traffic go out at full speed .... > > That because i dont want to shape my ssh connection & other connection from > my pc to bandwidth.limiter (:Then it''s better to create a separate class for it and put your ssh traffic in it.> I am still learning about fw mark to shape both incoming and outgoing > bandwidth from my network.Incoming traffic can be tricky. You need the imq device or a dedicated shaper for it so you can shape on 2 nic''s.> Please let me know how to shape incoming rate with Iptables fw mark > I ve done with the outgoing bandwidth but i still dont understand how to > mark incoming bandwidth from my ISP. > example: > > /usr/sbin/iptables -t mangle -A POSTROUTING -s 192.168.1.0/24 -j MARK > --set-mark 100 > this is for upgoing connection from my local network. > > tc class add dev eth0 parent 1: classid 1:100 htb rate 40Kbit ceil 40Kbit > tc qdisc add dev eth0 parent 1:100 handle 100 sfq perturb 10 quantum 1500 > tc filter add dev eth0 parent 1:0 protocol ip prio 200 handle 100 fw > classid 1:100The problem is not marking with iptables, but shaping. If you add something to nic eth0, you shape outgoing packets for that nic. You can not do this for incoming packets. What you can do is use imq. This is a virutal device and you can redirect incoming packets to it. And you can add a htb qdisc to it .... Or you have to use filters + policers, but that''s not powerful.> > Argh. Remove the burst and cburst parameters so htb can calculate the > > smalles one for you. I think that''s your problem. > > if i remove those parameters, it will exceed about 4 - 5 Kbit > if i use those parameters, it only exceed about 40-80 bit (:The exceed is normal. Traffic is bursty and you can not expect to shape it perfectly.> About my ISP, > Downstream connection is burstable. I get CIR 512Kbps burstable to 1024Kbps > Please advise how to deal this situation Stef,Then you have a problem. You can not configure this with htb because you don''t know the ISP bandwidth. However, you can create a setup for 1024 kbps. You will loose some precision because you are not the bottleneck anymore, but you will able to shape your link. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Tuesday 29 July 2003 15:34, Stef Coene wrote:> Then it''s better to create a separate class for it and put your ssh traffic > in it.okay, i ''ll do it your way ..> > I am still learning about fw mark to shape both incoming and outgoing > > bandwidth from my network. > Incoming traffic can be tricky. You need the imq device or a dedicated > shaper for it so you can shape on 2 nic''s.for marking outgoing from local network, i put those rules in POSTROUTING. i just want to know how about marking for downstream connection (from internet to my local network), where should i put this rule ? Note that i use different router for incoming and outgoing connection (multipath routing)..> Then you have a problem. You can not configure this with htb because you > don''t know the ISP bandwidth. However, you can create a setup for 1024 > kbps. You will loose some precision because you are not the bottleneck > anymore, but you will able to shape your link.Just a thought in my mind, script work ? latency check ? i need to figure this out, because all local ISP seems to be applied this kind of rule to make price for Internet bandwidth more cheaper .. And sadly i must take this ISP service because i cannot afford higher and clear channel bandwidth ratio. Regards, Rio Martin. -- Hell hath no fury like a bureaucrat scorned. -- Milton Friedman _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Tuesday 29 July 2003 10:57, Rio Martin. wrote:> On Tuesday 29 July 2003 15:34, Stef Coene wrote: > > Then it''s better to create a separate class for it and put your ssh > > traffic in it. > > okay, i ''ll do it your way .. > > > > I am still learning about fw mark to shape both incoming and outgoing > > > bandwidth from my network. > > > > Incoming traffic can be tricky. You need the imq device or a dedicated > > shaper for it so you can shape on 2 nic''s. > > for marking outgoing from local network, i put those rules in POSTROUTING. > i just want to know how about marking for downstream connection (from > internet to my local network), where should i put this rule ? Note that i > use different router for incoming and outgoing connection (multipath > routing)..It depends on how you want to mark, if you are natting, make sure you mark after the demasqueing so you know the real dst ip address.> > Then you have a problem. You can not configure this with htb because you > > don''t know the ISP bandwidth. However, you can create a setup for 1024 > > kbps. You will loose some precision because you are not the bottleneck > > anymore, but you will able to shape your link. > > Just a thought in my mind, script work ? latency check ? > i need to figure this out, because all local ISP seems to be applied this > kind of rule to make price for Internet bandwidth more cheaper ..Mhh. Even if you don''t know the bandwidth, you still can shape. It''s better to shape something then you don''t shape. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 30 July 2003 00:28, Stef Coene wrote:> > Just a thought in my mind, script work ? latency check ? > > i need to figure this out, because all local ISP seems to be applied this > > kind of rule to make price for Internet bandwidth more cheaper .. > Mhh. Even if you don''t know the bandwidth, you still can shape. It''s > better to shape something then you don''t shape.Thats why we should find out how much ISP give with script.. Perhaps we should ask for ISP help to put some file that we able to download all the time (just for bandwidth checker). Those script always check every 1 maybe 2 or 5 minutes to check how much bandwidth left and restart htb with the new bandwidth calculation. Btw , thats only on my mind, perhaps it would be to hard to implement anyway.. Regards, Rio Martin. -- There''s a fine line between courage and foolishness. Too bad it''s not a fence. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/