[LARTC] (no subject)
Hello every body, I install Traffic control on the my system(Kernel 2.4.7),when I adjust rule on that,my network rate decrease,I don't know why, In the normal operation I get 98Mbps ,but when I use cbq class ,and set the bandwith 50Mbit I get 15Mbps, :( what is the TC buttleneck ? I am really wait for you help, Thanks Noroozi - This message was sent using sharif web-based mail. http://mehr.sharif.edu ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Re: [LARTC] (no subject)
Le dim 18/08/2002 à 13:28, [EMAIL PROTECTED] a écrit : In the normal operation I get 98Mbps ,but when I use cbq class ,and set the bandwith 50Mbit I get 15Mbps, :( what is the TC buttleneck ? The bandwith parameter has to be set to the physical bandwith (use in internal computation), not to the deisirated one, use rate to specify the bandwith of the class. -- Eric Leblond Mail: [EMAIL PROTECTED] Tips and Linux: http://www.regit.org *---* Computers are like air conditioners. They don't work when Windows are open. *---* ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
[LARTC] Announce: Backport QoS from 2.4 to 2.2
Hello, I'm happy to announce a QoS Backport from 2.4 to 2.2. This is unofficial DiffServ v9 release (a.k.a DS-9). This beta release includes some bugfixes after the last 2.2 release, known as DS-8, many files copied from 2.4.19, some definitions and functions useful for keeping the interface for schedulers and classifiers similar to the 2.4 one. I'm planning to apply all 2.4/2.5 QoS fixes to this DS-9 tree, so the result should be fully functional net schedulers for 2.2. As there are still some 2.2 specific parts, any kind of reports are useful (even the good ones :)). Any bugfixes and comments can be posted directly to me or to the LARTC and/or DiffServ mailing lists. Have fun! DS-9 - QoS Backport from 2.4.19 to 2.2.21: http://www.ssi.bg/~ja/ds9/ Changes between DS-8 and DS-9: http://www.ssi.bg/~ja/ds9/Changelog.ds9 Regards -- Julian Anastasov [EMAIL PROTECTED] ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Re: [LARTC] HTTP traffic limiting
On Tue, 13 Aug 2002, Alexey Sheshka wrote: Hello! Sorry for ugly english. /sbin/iptables -A OUTPUT -t mangle -s x.x.x.x --sport 80 -j MARK --set-mark 199 And where is your tc filter statement? You must set some filters to catch the traffic into corresponding class. But it doesn't work ! Users download files via HTTP at full 100Mb, but if replase last line with /sbin/iptables -A OUTPUT -t mangle -s x.x.x.x -j MARK --set-mark 199 they got 51Mbit. Consider setting some TOS values, it could speed up some traffic if the routers on the route are configured correctly. You could then attach the traffic with each TOS to any of your HTB classes with tc filter statement. -- ## # | p0wer| # # __ |GG#1877248 | # # (oo) | [EMAIL PROTECTED] | # # / \/ \ Go away or I will replace you # # `V__V' with a very small shell script. # ## ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
[LARTC] Long delays
Hi. I'm using HTB to divide my 115kbit (95kbit of actual speed - HDSL) link for 10 people. IT seems to work fine for me. I have created such tree: ISP --, 1:1 - rate 95kbit, ceil 95kbit, burst 10k +--1:10 - rate 95kbit, ceil 95kbit, burst 10k, prio 0 | +--1:20 - rate 30kbit, ceil 40kbit, burst 2k, prio 1 | +--1:30 - rate 30kbit, ceil 40kbit, prio 3 | `--1:99 - rate 40kbit, ceil 50kbit, burst 5k, prio 2 Filters based on TOS (packets are mangled by firewall): - 1:10 - interactive class, low delays: SSH, telnet and some network games (like MoH,Q3,Battle.NET) - 1:20 - high reliability: WWW, mail, news... - 1:30 - high throughput: KaZaA, FTP... - 1:99 - default class All classes get also the SFQ. The problem is that when classes other than 1:10 are loaded, even slightly, the delays on 1:10 grow too large, thus distorting heavily the network games. When that classes are very heavily loaded, no reasonable playing is possible. I want to reduce the delays in 1:10 to the absolute minimum, especially when the other classes aren't heavily loaded. What should I do? Too slow CPU (P166MMX) ? Greetz. PS. What is the unit for burst: bit/s, byte/s, packet/s or something else? -- ## # | p0wer| # # __ |GG#1877248 | # # (oo) | [EMAIL PROTECTED] | # # / \/ \ Go away or I will replace you # # `V__V' with a very small shell script. # ## ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Re: [LARTC] Another sharing tehnique, is this possible ?
Say I have a internet link with 100kbits bandwith, then I want to share it between many clients (which will increase over time). Let's i start with 5 clients with rate = 30kbits... See the total bandwith of users is 120kbits but I have only 100kbits So where is the problem I want to describe their bandwith like total 100kbits |__ user1 = rate 30kbits, ceil 30kbits |__ user2 = rate 30kbits, ceil 30kbits |__ user3 = rate 30kbits, ceil 30kbits |__ . |__ userX = rate 30kbits, ceil 30kbits U can say why I just don't calculate 100/5 = 20 and set for all users rate 20kbits, ceil 30kbits, 'cause i will add more users in the future and will need to calculate again and again this value... also some of them may want 30kbits other 10kbits etc (it is ok for them to get lower rates 'cause the speed is ungaranteed the user are not online 24hours a day and of course when bandwith got used to the max it will upgraded to better speed).. In fact what I want to say is : the max rate should be X but if the link is overused u will get lower speed ...?? One way this to be done is if I use something like this : rate 0 ceil desired-speed but is this possible, or if not what is the lowest possible value - Will proirity have be taken into account so that i can say some should be served better than other ? - the lower possible bandwith i will lend is 9600bps .. yes we have such speeds the price here is big :( Normally, each class gets the rate as a minimum. After that, the remaning bandwidth is divided according to the proportions of the rate (more accurate, the quantums) but the classes with the lowest prio will be serverd first. So only remaining bandwidth will be influenced with the prio parameter. And giving a class rate 0 is a strange idea :) Stef -- [EMAIL PROTECTED] Using Linux as bandwidth manager http://www.docum.org/ #lartc @ irc.oftc.net ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Re: [LARTC] HTB equivalent for 'bounded' and 'isolated' in CBQ
On Friday 16 August 2002 23:09, Amit Kucheria wrote: Hi all, Sally Floyd's Link sharing paper defines the following: 1. Bounded class: class that is not allowed to borrow from ancestor classes, regardless of the limit status of those classes. 2. Isolated class: class that does not allow non-descendant classes to borrow its unused bandwidth and that does not borrow bandwidth from other classes in turn How does HTB implement a 'bounded class'? I seem to see no constructs to be able to do this. You can use the ceil parameter to bound a class to a certain rate. With rate = ceil, you will get the same result as the bounded option in cbq. The 'ceil' parameter allows us to implement a part of the 'isolated' definition above. By setting it equal to 'rate', the class does not borrow bandwidth, but what will stop other classes from borrowing its unused bandwidth? Providing no ceil parameter means rate = ceil. But this can not be used to implement isolated like in cbq. Rate = ceil means the class can not use more bandwidth then it's rate. Isolated in cbq means other classes can not borrow bandwidth from the class and that's not the same. There is no way you can implement isolated with htb. But with htb you can do something like this : Total : 100 class1 rate 20 ceil 20 class2 rate 40 ceil 80 class3 rate 40 ceil 80 class 1 is isolated like in cbq. It can not use more then it's rate/ceil and class2 and class3 will never use bandwidth from class1, only from each other. Just like the definition of isolated :) Stef Btw, I never could get isolated working with cbq. -- [EMAIL PROTECTED] Using Linux as bandwidth manager http://www.docum.org/ #lartc @ irc.oftc.net ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Re: [LARTC] Another sharing tehnique, is this possible ?
On Sunday 18 August 2002 21:56, Ivo De Decker wrote: On Sun, Aug 18, 2002 at 09:48:09PM +0200, Stef Coene wrote: Normally, each class gets the rate as a minimum. After that, the remaning bandwidth is divided according to the proportions of the rate (more accurate, the quantums) but the classes with the lowest prio will be serverd first. So only remaining bandwidth will be influenced with the prio parameter. And giving a class rate 0 is a strange idea :) Well, not really: There might be some sorts of traffic (eg backups) that aren't really prioritary. The classes of this traffic could use borrowed bandwith exclusively: if there is no other traffic (or not enough to fill the total available bandwith), the backups can happen, but if there is other (higher priority) traffic, the backups can wait till that traffic lessens (assuming this eventually happens, off course). The question is: is this possible? If you have a total of 100, give backup 1 and the rest 99. As long as there is no other traffic, backup can get the full bandwidth. But if there is other traffic, it can use up to 99. Stef -- [EMAIL PROTECTED] Using Linux as bandwidth manager http://www.docum.org/ #lartc @ irc.oftc.net ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
[LARTC] filter in-cache traffic with tc
Hello lartc, I am using squid in transparent mode, and bandwith control with tc (htb). (redhat 7.3). Currently I mark packets comming from 3128 (squid port) to users, there are any way to filter the trafic if the object that user request alreay is in the squid cache (hit)? sorry for my english. -- Best regards, Fabian mailto:[EMAIL PROTECTED] ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
[LARTC] Re: LARTC digest, Vol 1 #723 - 10 msgs
Text Item Description: Binary data I will be out of the office from 8/19 until 8/23. I will be returning on Monday afternoon, 8/26, and will reply then. Thank you, Michael Pellegrino Softerware, Inc. ___ Subject: LARTC digest, Vol 1 #723 - 10 msgs From: [EMAIL PROTECTED] Date: 08/18/02 21:26:02 This item was automatically created and contains MIME Information.
[LARTC] VoIP and QoS
Title: VoIP and QoS There are two Linux routers, a 64K leased line is used to connect these two machines. Now, we would like to run Voice over IP between two machines. We have to run QoS and tag voice packet's TOS field, otherwise, the voice quality is terrible if we transfer file and make voice call at the same time. I am not familiar with tc in Linux. But I have copied a script from LARTC archives. If the following config is used, the situation is better than nothing used. The voice packets can be tagged as TOS=0x21: tc qdisc add dev hdlc0 parent 1:1 handle 2: pfifo limit 512 tc qdisc add dev hdlc0 parent 1:2 handle 3: pfifo limit 512 tc qdisc add dev hdlc0 parent 1:3 handle 4: pfifo limit 512 tc filter add dev hdlc0 parent 1:0 protocol ip prio 1 u32 divisor 1 tc filter add dev hdlc0 parent 1:0 prio 1 u32 match ip tos 0x21 0xff flowid 1:1 I don't know what it means exactly. But this is used, the voice quality is bad at the beginning when file is transferred. The voice quality will converge to be good after about 30 seconds. However, this is not acceptable. Can anyone tell me what is the best tc config for running VoIP? Thanks very much.
Re: [LARTC] filter in-cache traffic with tc
Hello Fabian, I recomend you use squid delay pools or WRR qdisc with patch for squid. see this links http://wipl-wrr.sourceforge.net/ http://squid-docs.sourceforge.net/latest/html/x2116.html Monday, August 19, 2002, 7:31:05 AM, you wrote: FG Hello lartc, I am using squid in transparent mode, and bandwith control FG with tc (htb). (redhat 7.3). FG Currently I mark packets comming from 3128 (squid FG port) to users, there are any way to filter the trafic if the object FG that user request alreay is in the squid cache (hit)? FG sorry for my english. --- mailto:[EMAIL PROTECTED] BR Alexey Talikov FORTEK --- ___ LARTC mailing list / [EMAIL PROTECTED] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
[LARTC] Re: LARTC digest, Vol 1 #725 - 2 msgs
I will be out of the office from 8/19 until 8/23. I will be returning on Monday afternoon, 8/26, and will reply then. Thank you, Michael Pellegrino Softerware, Inc. ___ Subject: LARTC digest, Vol 1 #725 - 2 msgs From: [EMAIL PROTECTED] Date: 08/19/02 01:27:39 Send LARTC mailing list submissions to [EMAIL PROTECTED] To subscribe or unsubscribe via the World Wide Web, visit http://mailman.ds9a.nl/mailman/listinfo/lartc or, via email, send a message with subject or body 'help' to [EMAIL PROTECTED] You can reach the person managing the list at [EMAIL PROTECTED] When replying, please edit your Subject line so it is more specific than Re: Contents of LARTC digest... Today's Topics: 1. VoIP and QoS (Patrick Chan) 2. Re: filter in-cache traffic with tc (Alexey Talikov) --__--__-- Message: 1 From: Patrick Chan [EMAIL PROTECTED] To: [EMAIL PROTECTED] Date: Mon, 19 Aug 2002 10:01:18 +0800 Subject: [LARTC] VoIP and QoS This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. --_=_NextPart_001_01C24724.4EFA3390 Content-Type: text/plain; charset=iso-8859-1 There are two Linux routers, a 64K leased line is used to connect these two machines. Now, we would like to run Voice over IP between two machines. We have to run QoS and tag voice packet's TOS field, otherwise, the voice quality is terrible if we transfer file and make voice call at the same time. I am not familiar with tc in Linux. But I have copied a script from LARTC archives. If the following config is used, the situation is better than nothing used. The voice packets can be tagged as TOS=0x21: tc qdisc add dev hdlc0 parent 1:1 handle 2: pfifo limit 512 tc qdisc add dev hdlc0 parent 1:2 handle 3: pfifo limit 512 tc qdisc add dev hdlc0 parent 1:3 handle 4: pfifo limit 512 tc filter add dev hdlc0 parent 1:0 protocol ip prio 1 u32 divisor 1 tc filter add dev hdlc0 parent 1:0 prio 1 u32 match ip tos 0x21 0xff flowid 1:1 I don't know what it means exactly. But this is used, the voice quality is bad at the beginning when file is transferred. The voice quality will converge to be good after about 30 seconds. However, this is not acceptable. Can anyone tell me what is the best tc config for running VoIP? Thanks very much. --_=_NextPart_001_01C24724.4EFA3390 Content-Type: text/html; charset=iso-8859-1 !DOCTYPE HTML PUBLIC -//W3C//DTD HTML 3.2//EN HTML HEAD META HTTP-EQUIV=Content-Type CONTENT=text/html; charset=iso-8859-1 META NAME=Generator CONTENT=MS Exchange Server version 5.5.2653.12 TITLEVoIP and QoS/TITLE /HEAD BODY PFONT SIZE=2There are two Linux routers, a 64K leased line is used/FONT BRFONT SIZE=2to connect these two machines. Now, we would like to run/FONT BRFONT SIZE=2Voice over IP between two machines. We have to run QoS and /FONT BRFONT SIZE=2tag voice packet's TOS field,/FONT BRFONT SIZE=2otherwise, the voice quality is terrible if we transfer file/FONT BRFONT SIZE=2and make voice call at the same time./FONT /P PFONT SIZE=2I am not familiar with tc in Linux. But I have copied a script/FONT BRFONT SIZE=2from LARTC archives. If the following config is used, the situation/FONT BRFONT SIZE=2is better than nothing used. The voice packets can be tagged as /FONT BRFONT SIZE=2TOS=0x21:/FONT /P PFONT SIZE=2tc qdisc add dev hdlc0 parent 1:1 handle 2: pfifo limit 512/FONT BRFONT SIZE=2tc qdisc add dev hdlc0 parent 1:2 handle 3: pfifo limit 512/FONT BRFONT SIZE=2tc qdisc add dev hdlc0 parent 1:3 handle 4: pfifo limit 512/FONT BRFONT SIZE=2tc filter add dev hdlc0 parent 1:0 protocol ip prio 1 u32 divisor 1/FONT BRFONT SIZE=2tc filter add dev hdlc0 parent 1:0 prio 1 u32 match ip tos 0x21 0xff flowid 1:1/FONT /P PFONT SIZE=2I don't know what it means exactly. But this is used, the voice quality/FONT BRFONT SIZE=2is bad at the beginning when file is transferred. The voice quality will/FONT BRFONT SIZE=2quot;convergequot; to be good after about 30 seconds./FONT /P PFONT SIZE=2However, this is not acceptable. Can anyone tell me what is the best/FONT BRFONT SIZE=2tc config for running VoIP? Thanks very much./FONT /P /BODY /HTML --_=_NextPart_001_01C24724.4EFA3390-- --__--__-- Message: 2 Date: Mon, 19 Aug 2002 09:05:12 +0500 From: Alexey Talikov [EMAIL PROTECTED] Reply-To: Alexey Talikov [EMAIL PROTECTED] Organization: Fortek To: Fabian Gervan [EMAIL PROTECTED] Cc: [EMAIL PROTECTED] Subject: Re: [LARTC] filter in-cache traffic with tc Hello Fabian, I recomend you use squid delay pools or WRR qdisc with patch for squid. see this links http://wipl-wrr.sourceforge.net/ http://squid-docs.sourceforge.net/latest/html/x2116.html Monday, August 19, 2002, 7:31:05 AM, you wrote: FG Hello lartc, I am using squid in transparent mode, and bandwith control FG with tc (htb). (redhat 7.3). FG Currently I mark packets comming from 3128
