Hello All, I aa new subscribin this forum, and have worked with netem in recent months. I need sominfo and help fropeople who know how we can generate and put thdistribution tablin /usr/lib/tc. BecausI wanto emulate a case, that some random spikes of large delay(e.g. 500ms) occurwhen thoverall delay is a small value(4ms), and thprobabolty of spikes largdelays is very small , I tried pareto dis, buit didnt work. (The large spikes were not of the order thaI wish to be) Thanks, maliheh -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20080404/7015d286/attachment.ht Fromaliheh.yousefi aalumni.chalmers.se Fri Apr 4 00:34:49 2008 From: maliheh.yousefi aalumni.chalmers.s(Maliheh Yousefi) Date: Fri, 4 Apr 2008 09:34:49 +0200 Subject: Generating DistributioTable Message-ID: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> OFri, Apr 4, 2008 a9:32 AM, Maliheh Yousefi < maliheh.yousefi aalumni.chalmers.se> wrote: > > Hello All, > I aa new subscribin this forum, and have worked with netem in recent > months. > I need sominfo and help fropeople who know how we can generate and put > thdistribution tablin /usr/lib/tc. > BecausI wanto emulate a case, that some random spikes of large > delay(e.g. 500ms) occurwhen thoverall delay is a small value(4ms), and > thprobability of largdelay spikes is very small , > > I tried pareto dis, buit didnt work. (The large spikes were not of the > order thaI wish to be) > > Thanks, > maliheh > -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20080404/88b0d4a8/attachment.ht Froshemminger alinux-foundation.org Fri Apr 4 08:18:17 2008 From: shemminger alinux-foundation.org (Stephen Hemminger) Date: Fri, 4 Apr 2008 08:18:17 -0700 Subject: Generating DistributioTable In-Reply-To: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> References: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> Message-ID: <20080404081817.3b474378@extreme> OFri, 4 Apr 2008 09:34:49 +0200 "Maliheh Yousefi" <maliheh.yousefi aalumni.chalmers.se> wrote: > OFri, Apr 4, 2008 a9:32 AM, Maliheh Yousefi < > maliheh.yousefi aalumni.chalmers.se> wrote: > > > > > Hello All, > > I aa new subscribin this forum, and have worked with netem in recent > > months. > > I need sominfo and help fropeople who know how we can generate and put > > thdistribution tablin /usr/lib/tc. > > BecausI wanto emulate a case, that some random spikes of large > > delay(e.g. 500ms) occurwhen thoverall delay is a small value(4ms), and > > thprobability of largdelay spikes is very small , > > > > I tried pareto dis, buit didnt work. (The large spikes were not of the > > order thaI wish to be) > > > > Thanks, > > maliheh > > Look aMakefiland input data in the iproute2 source. Theris a tool therto convert a set of data points to a distribution table. Note: thcommand lindoesn't have a hard coded list of tables so if you wanyou can do: maketablsome_samples.da>/usr/lib/tc/spiky2.dist and theusspiky2 as an option on the command line. Frojuliokriger agmail.com Thu Apr 10 17:16:39 2008 From: juliokriger agmail.co(Julio Kriger) Date: Thu, 10 Apr 2008 21:16:39 -0300 Subject: emulatabrupchanges on RTT Message-ID: <682bc30a0804101716l4516b6cdy5e88585fecaf3d95@xxxxxxxxxxxxxx> Hi! I hav3 PC's. 2 of theemulate end-points, the other one acts as bridge. How do I havto senetem (on the brifge PC) to emulate abrupt changes on RTT? TIA. Regards, Julio -- -- Frobogus@xxxxxxxxxxxxxxxxx Thu Apr 10 09:22:30 2008 From: bogus@xxxxxxxxxxxxxxxxx() Date: Thu, 10 Apr 2008 16:22:30 -0000 Subject: No subject Message-ID: <mailman.3.1207873008.19971.netem@xxxxxxxxxxxxxxxxxxxxxxxxxx> convulsed with laughter. Someday I intend reading it. Groucho Marx ---------------------------- Julio Kriger mailto:juliokriger agmail.com Fromaliheh.yousefi aalumni.chalmers.se Wed Apr 16 08:43:53 2008 From: maliheh.yousefi aalumni.chalmers.s(Maliheh Yousefi) Date: Wed, 16 Apr 2008 17:43:53 +0200 Subject: Generating DistributioTable In-Reply-To: <20080404081817.3b474378@extreme> References: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> <20080404081817.3b474378@extreme> Message-ID: <d2a38fe70804160843y7af900d5wea258e4caa7fac40@xxxxxxxxxxxxxx> Hi Stephen, thanks for your prompreply, I did apply thTCsolution, (trace), and kernel iproute2 patch. it did work well. Now I wonder how to apply firstly a tracfiland secondly a tbf filter to all thpkts exiting eth0. . iNete ( http://www.linux-foundation.org/en/Net:Netem ), i saw these commands: # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limi3000 # tc -s qdisc ls dev eth0 qdisc nete1: limi1000 delay 100.0ms Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) qdisc tbf 10: rat256Kbiburst 1599b lat 26.6ms Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) which afirsa constant delay is applied and then a tbf,....exactly my case,excepthaI wanna apply my trace file for delays in root. BUT, this solutiodoesnwork, mean that only the child rule is applied to th pkts (tbf), nothe delay on the root,.... so I awondering is iat all possible to apply different qdisc to pkts exiting ainterface? or I was wrong isetting sth ...? BesRegards, Maliheh OFri, Apr 4, 2008 a5:18 PM, Stephen Hemminger < shemminger alinux-foundation.org> wrote: > OFri, 4 Apr 2008 09:34:49 +0200 > "Maliheh Yousefi" <maliheh.yousefi aalumni.chalmers.se> wrote: > > > OFri, Apr 4, 2008 a9:32 AM, Maliheh Yousefi < > > maliheh.yousefi aalumni.chalmers.se> wrote: > > > > > > > > Hello All, > > > I aa new subscribin this forum, and have worked with netem in > recent > > > months. > > > I need sominfo and help fropeople who know how we can generate and > put > > > thdistribution tablin /usr/lib/tc. > > > BecausI wanto emulate a case, that some random spikes of large > > > delay(e.g. 500ms) occurwhen thoverall delay is a small value(4ms), > and > > > thprobability of largdelay spikes is very small , > > > > > > I tried pareto dis, buit didnt work. (The large spikes were not of > the > > > order thaI wish to be) > > > > > > Thanks, > > > maliheh > > > > > Look aMakefiland input data in the iproute2 source. > Theris a tool therto convert a set of data points to a distribution > table. > Note: thcommand lindoesn't have a hard coded list of tables so > if you wanyou can do: > > maketablsome_samples.da>/usr/lib/tc/spiky2.dist > > and theusspiky2 as an option on the command line. > -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20080416/3d382f37/attachment.ht Fromaliheh.yousefi aalumni.chalmers.se Thu Apr 17 05:04:46 2008 From: maliheh.yousefi aalumni.chalmers.s(Maliheh Yousefi) Date: Thu, 17 Apr 2008 14:04:46 +0200 Subject: Generating DistributioTable In-Reply-To: <d2a38fe70804160843y7af900d5wea258e4caa7fac40@xxxxxxxxxxxxxx> References: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> <20080404081817.3b474378@extreme> <d2a38fe70804160843y7af900d5wea258e4caa7fac40@xxxxxxxxxxxxxx> Message-ID: <d2a38fe70804170504w1fc0acc9he708ffe9201e8f29@xxxxxxxxxxxxxx> Hi again, I found oumy problewith the previous case,.... so no more problems there,. buI wonder how kernel treats two consequenidentical netem commands? like: tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: netem delay 50ms does idelay each pk, 150ms or only the child qdisc? which is 50ms ? My ping tests show only thlasone is applied, bus is that really the case? BesRegards, Maliheh OWed, Apr 16, 2008 a5:43 PM, Maliheh Yousefi < maliheh.yousefi aalumni.chalmers.se> wrote: > > Hi Stephen, > thanks for your prompreply, > I did apply thTCsolution, (trace), and kernel iproute2 patch. it did > work well. > Now I wonder how to apply firstly a tracfiland secondly a tbf filter > to all thpkts exiting eth0. . > > iNete ( http://www.linux-foundation.org/en/Net:Netem ), i saw these > commands: > > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 > # tc -s qdisc ls dev eth0 > qdisc nete1: limi1000 delay 100.0ms > Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) > qdisc tbf 10: rat256Kbiburst 1599b lat 26.6ms > > Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) > > > which afirsa constant delay is applied and then a tbf,....exactly my > case,excepthaI wanna apply my trace file for delays in root. > BUT, this solutiodoesnwork, mean that only the child rule is applied > to th pkts (tbf), nothe delay on the root,.... > so I awondering is iat all possible to apply different qdisc to pkts > exiting ainterface? > or I was wrong isetting sth ...? > > BesRegards, > Maliheh > > > > OFri, Apr 4, 2008 a5:18 PM, Stephen Hemminger < > shemminger alinux-foundation.org> wrote: > > > OFri, 4 Apr 2008 09:34:49 +0200 > > "Maliheh Yousefi" <maliheh.yousefi aalumni.chalmers.se> wrote: > > > > > OFri, Apr 4, 2008 a9:32 AM, Maliheh Yousefi < > > > maliheh.yousefi aalumni.chalmers.se> wrote: > > > > > > > > > > > Hello All, > > > > I aa new subscribin this forum, and have worked with netem in > > recent > > > > months. > > > > I need sominfo and help fropeople who know how we can generate > > and put > > > > thdistribution tablin /usr/lib/tc. > > > > BecausI wanto emulate a case, that some random spikes of large > > > > delay(e.g. 500ms) occurwhen thoverall delay is a small > > value(4ms), and > > > > thprobability of largdelay spikes is very small , > > > > > > > > I tried pareto dis, buit didnt work. (The large spikes were not > > of the > > > > order thaI wish to be) > > > > > > > > Thanks, > > > > maliheh > > > > > > > > Look aMakefiland input data in the iproute2 source. > > Theris a tool therto convert a set of data points to a distribution > > table. > > Note: thcommand lindoesn't have a hard coded list of tables so > > if you wanyou can do: > > > > maketablsome_samples.da>/usr/lib/tc/spiky2.dist > > > > and theusspiky2 as an option on the command line. > > > > -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20080417/85c11380/attachment.ht Frojuliokriger agmail.com Thu Apr 17 06:00:15 2008 From: juliokriger agmail.co(Julio Kriger) Date: Thu, 17 Apr 2008 10:00:15 -0300 Subject: Generating DistributioTable In-Reply-To: <d2a38fe70804170504w1fc0acc9he708ffe9201e8f29@xxxxxxxxxxxxxx> References: <d2a38fe70804040034n497647bgf7b6f388cc158aab@xxxxxxxxxxxxxx> <20080404081817.3b474378@extreme> <d2a38fe70804160843y7af900d5wea258e4caa7fac40@xxxxxxxxxxxxxx> <d2a38fe70804170504w1fc0acc9he708ffe9201e8f29@xxxxxxxxxxxxxx> Message-ID: <682bc30a0804170600v58feee11pd63924773ec0b38e@xxxxxxxxxxxxxx> Hi! Netedoesn'support nest with it self. If you wana "fixed delay" plus thusual delay options of netem, I hava patch to do that. Regards, Julio O4/17/08, Maliheh Yousefi <maliheh.yousefi aalumni.chalmers.se> wrote: > Hi again, > I found oumy problewith the previous case,.... so no more problems > there,. > buI wonder how kernel treats two consequenidentical netem commands? > like: > tc qdisc add dev eth0 roohandl1: netem delay 100ms > tc qdisc add dev eth0 paren1:1 handl10: netem delay 50ms > > does idelay each pk, 150ms or only the child qdisc? which is 50ms ? > My ping tests show only thlasone is applied, bus is that really the > case? > > BesRegards, > Maliheh > > > > OWed, Apr 16, 2008 a5:43 PM, Maliheh Yousefi > <maliheh.yousefi aalumni.chalmers.se> wrote: > > > > Hi Stephen, > > thanks for your prompreply, > > I did apply thTCsolution, (trace), and kernel iproute2 patch. it did > work well. > > Now I wonder how to apply firstly a tracfiland secondly a tbf filter > to all thpkts exiting eth0. . > > > > iNete ( http://www.linux-foundation.org/en/Net:Netem > ), i saw thescommands: > > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer > 1600 limi3000 > > # tc -s qdisc ls dev eth0 > > qdisc nete1: limi1000 delay 100.0ms > > Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) > > qdisc tbf 10: rat256Kbiburst 1599b lat 26.6ms > > Sen0 bytes 0 pkts (dropped 0, overlimits 0 ) > > > > which afirsa constant delay is applied and then a tbf,....exactly my > case,excepthaI wanna apply my trace file for delays in root. > > BUT, this solutiodoesnwork, mean that only the child rule is applied > to th pkts (tbf), nothe delay on the root,.... > > so I awondering is iat all possible to apply different qdisc to pkts > exiting ainterface? > > or I was wrong isetting sth ...? > > > > BesRegards, > > Maliheh > > > > > > > > > > > > > > OFri, Apr 4, 2008 a5:18 PM, Stephen Hemminger > <shemminger alinux-foundation.org> wrote: > > > > > OFri, 4 Apr 2008 09:34:49 +0200 > > > > > > > > > > > > "Maliheh Yousefi" <maliheh.yousefi aalumni.chalmers.se> > wrote: > > > > > > > OFri, Apr 4, 2008 a9:32 AM, Maliheh Yousefi < > > > > maliheh.yousefi aalumni.chalmers.se> wrote: > > > > > > > > > > > > > > Hello All, > > > > > I aa new subscribin this forum, and have worked with netem in > recent > > > > > months. > > > > > I need sominfo and help fropeople who know how we can generate > and put > > > > > thdistribution tablin /usr/lib/tc. > > > > > BecausI wanto emulate a case, that some random spikes of large > > > > > delay(e.g. 500ms) occurwhen thoverall delay is a small > value(4ms), and > > > > > thprobability of largdelay spikes is very small , > > > > > > > > > > I tried pareto dis, buit didnt work. (The large spikes were not > of the > > > > > order thaI wish to be) > > > > > > > > > > Thanks, > > > > > maliheh > > > > > > > > > > > Look aMakefiland input data in the iproute2 source. > > > Theris a tool therto convert a set of data points to a distribution > table. > > > Note: thcommand lindoesn't have a hard coded list of tables so > > > if you wanyou can do: > > > > > > maketablsome_samples.da>/usr/lib/tc/spiky2.dist > > > > > > and theusspiky2 as an option on the command line. > > > > > > > > > > _______________________________________________ > Netemailing list > Netealists.linux-foundation.org > https://lists.linux-foundation.org/mailman/listinfo/netem > -- -- Frobogus@xxxxxxxxxxxxxxxxx Wed Apr 16 13:51:24 2008 From: bogus@xxxxxxxxxxxxxxxxx() Date: Wed, 16 Apr 2008 20:51:24 -0000 Subject: No subject Message-ID: <mailman.0.1208437249.30820.netem@xxxxxxxxxxxxxxxxxxxxxxxxxx> convulsed with laughter. Someday I intend reading it. Groucho Marx ---------------------------- Julio Kriger mailto:juliokriger agmail.com Frocubic1271 agmail.com Thu Apr 24 23:29:04 2008 From: cubic1271 agmail.co(cubic1271 1271) Date: Fri, 25 Apr 2008 02:29:04 -0400 Subject: Neteissuin Ubuntu 7.10 / Fedora 5,6 Message-ID: <367782b60804242329u2216accdw52bd14b22d12b274@xxxxxxxxxxxxxx> Howdy folks. I skimmed thlisarchives, but didn't really see anything abouthis; if I missed something, feel freto tell me when / wheriwas and I can go look that up. Anyway, here goes: I had a brand new install of ubuntu 7.10: 'Linux mybox 2.6.22-14-generic #1 SMP SuOc14 23:05:12 GMT 2007 i686 GNU/Linux' oa 1.6 ghz P4 with 128 MB of RAM. I was trying to seup a few NICs with larg(read: 3 minute) delays for a small projecI'm working on. I decided to keep thdelays small afirst, just so I could test and maksureverything worked. To this end, I did: 'sudo tc qdisc add dev eth0 roonetedelay 4000ms' Which should'vadded a 4 second delay to every outbound packeon thainterface. On my box, I received thfollowing errors: myuser amybox:~$ sudo tc qdisc add dev eth0 roonetem delay 4000ms [sudo] password for myuser: Illegal 4000000000 tim(too large) Illegal "latency" I also tried doing 'sudo tc qdisc add dev eth0 roonetedelay 4s' which worked, buthinterface didn't delay correctly (only delayed for abou580 ms instead of 4 sec). Neteshowed similar behavior for mon FC5/6, though seemed to work finwhen I grabbed a version of Gentoo and used tha(with a build of vanilla 2.6.21). I'vsincthrown ubuntu 7.10 onto a few more boxes, with similar results. Much as I likGentoo, I really don'particularly like having to waifor -everything- to compil/ install itself just to gea tesbox set up (a few hours for a Gentoo installation vs. 30 minutes for aUbuntu installation). Thus, being ablto use Ubuntu and / or Fedora for thstuff I need would bmuch nicer than Gentoo. Thanks iadvancfor any enlightenment ya'll can provide! --GilberClark
