From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from blu0-omc1-s15.blu0.hotmail.com (blu0-omc1-s15.blu0.hotmail.com [65.55.116.26]) by huchra.bufferbloat.net (Postfix) with ESMTP id D172B21F1D4 for ; Fri, 31 Jan 2014 05:36:24 -0800 (PST) Received: from BLU178-W43 ([65.55.116.7]) by blu0-omc1-s15.blu0.hotmail.com with Microsoft SMTPSVC(6.0.3790.4675); Fri, 31 Jan 2014 05:36:23 -0800 X-TMN: [bsC/CfnchiWoipaZNc3FQTHLIcoh8aSl] X-Originating-Email: [jeremy_tourville@hotmail.com] Message-ID: Content-Type: multipart/alternative; boundary="_988dd7de-9b84-4693-a08f-04cd29912e0f_" From: Jeremy Tourville To: "cerowrt-users@lists.bufferbloat.net" Date: Fri, 31 Jan 2014 07:36:23 -0600 Importance: Normal In-Reply-To: References: MIME-Version: 1.0 X-OriginalArrivalTime: 31 Jan 2014 13:36:23.0973 (UTC) FILETIME=[6F105D50:01CF1E89] Subject: Re: [Cerowrt-users] Cerowrt-users Digest, Vol 13, Issue 2 X-BeenThere: cerowrt-users@lists.bufferbloat.net X-Mailman-Version: 2.1.13 Precedence: list List-Id: Support for user problems regarding cerowrt List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 31 Jan 2014 13:36:25 -0000 --_988dd7de-9b84-4693-a08f-04cd29912e0f_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable =0A= =0A= =0A= =0A= =0A= =0A= Sebastian and Dave=2C thanks to both of you for your email. I did some mo= re checking on settings in my modem. The max line rate is listed as 7616kb= ps down and 512 kbps up in the modem setup page. My ISP labels this as 6 M= B service. The actual throughput I was getting is 5700kbps and 450 kbps be= fore SQM. =20 >>>So you specified5.7 and 0.45? No=2C at the time I specified 5130 and 400 based on actual download speeds.= =20 I did some more testing this evening. Based on Dave's comment I upgraded t= o the latest build 3.10.26. I also adjusted my ingress and egress speeds t= o 6854 and 460. When I did this my speeds jumped back up to near normal. = So..... I think if I understand correctly you should set your speed to appr= oximately 90% of the listed line rate in your modem. > > What info do I need to get from my ISP to best optimize my connection? >=20 >=0A= The actual line rates for down- and uplink as well as the full =0A= encapsulation information: DHCP or PPPoE or PPPoA=3B >VC-MUX or LLC/SNAP=3B= =0A= VLAN or no vlan. My ISP does PPPoE (mine is done on the router instead of the modem - I'm in= bridged mode) Encapsulation is LLC. I also found out the interleave is s= et at 64. As I understand it=2C that has an affect on latency. Anyhow=2C = I think that should cover the most critical info needed. I currently have= my per packet overhead set at 44. =20 I'll have to dig in and read some more so I better understand what this set= ting does. I did take a brief look at the RRUL tool. I'll be teaching mys= elf how to set it up and use it so I can run some more thorough tests this = weekend.=20 >>(Note it is possible to empirically measure the overhead on your link=2C = =0A= and I would be happy to help you with this=2C just contact me if =0A= interested). Yes=2C I'd be interested in learning more! I think both of your emails have helped to clarify what my settings need to= be and more importantly I am better understanding HOW you arrived at the a= ppropriate values and the interplay between them. If I can help with docum= entation of the SQM setup page on the wiki please let me know. > From: cerowrt-users-request@lists.bufferbloat.net > Subject: Cerowrt-users Digest=2C Vol 13=2C Issue 2 > To: cerowrt-users@lists.bufferbloat.net > Date: Thu=2C 30 Jan 2014 12:00:01 -0800 >=20 > Send Cerowrt-users mailing list submissions to > cerowrt-users@lists.bufferbloat.net >=20 > To subscribe or unsubscribe via the World Wide Web=2C visit > https://lists.bufferbloat.net/listinfo/cerowrt-users > or=2C via email=2C send a message with subject or body 'help' to > cerowrt-users-request@lists.bufferbloat.net >=20 > You can reach the person managing the list at > cerowrt-users-owner@lists.bufferbloat.net >=20 > When replying=2C please edit your Subject line so it is more specific > than "Re: Contents of Cerowrt-users digest..." >=20 >=20 > Today's Topics: >=20 > 1. Re: SQM Setup and Performance (Sebastian Moeller) >=20 >=20 > ---------------------------------------------------------------------- >=20 > Message: 1 > Date: Thu=2C 30 Jan 2014 20:29:50 +0100 > From: Sebastian Moeller > To: Jeremy Tourville > Cc: "cerowrt-users@lists.bufferbloat.net" > > Subject: Re: [Cerowrt-users] SQM Setup and Performance > Message-ID: <646BCC1A-0C7A-46AB-827D-717AEA0A2174@gmx.de> > Content-Type: text/plain=3B charset=3Diso-8859-1 >=20 > Hi Jeremy=2C >=20 > On Jan 30=2C 2014=2C at 18:06 =2C Jeremy Tourville = wrote: >=20 > > Hello=2C I followed your excellent instructions here - > > http://www.bufferbloat.net/projects/cerowrt/wiki/Setting_up_SQM_for_Cer= oWrt_310 > > =20 > > I am using build 3.10.24-8 > > =20 > > I am using a DSL line rated at 6 Mbps down and 512kbps up. My real thr= oughput without SQM enabled is 5.7Mbps down and 450kbps up. >=20 > This looks interesting=2C the ATM 48 bytes in 53 byte cell encapsulation= used in ADSL only leaves 100*48/53 =3D 90.57% percent of the specified lin= e rate available for your traffic (but that still contains further per pack= et overhead). So I would assume that the actual liberate is >6.3Mbps? Do yo= u have any chance of verifying the line rate to the DSLAM? Many modems/dsl-= routers have offer a web page that gives some statistics and information li= ke the line rate. You will need the line rate as precise as possible if yo= u want to minimize the bandwidth "sacrifice" needed to keep latencies reaso= nable (if in doubt err on the too small side though). >=20 > > After enabling SQM my throughput has dropped to approximately 4.5Mbps d= own and 350 kbps up. =20 >=20 > So you specified5.7 and 0.45? Then the link layer adaptation mechanism w= ill try to account for the 48in53 problem and cut down your available rates= by ~10% so the best you can expect is ~5.1 and 0.4. If you specified 90% o= f the measured speed you are already at 5.7 * 0.9 (90% of measured) =3D 5.1= 3 * 0.9 (fixed ATM overhead) =3D 4.617 and 450 * 0.9 (90% of measured) =3D = 405 * 0.9 (fixed ATM overhead) =3D 364.5. So you can not rally expect much = more than you got. >=20 >=20 > > Does this seem like an amount that is expected? (within norms?) > > It would seem reasonable that I should expect some performance loss at = the expense of better bufferbloat management based on setting 85-95% of act= ual download/upload speeds. Please correct me if I am wrong. :-) >=20 > Now=2C my recommendation for ADSL links (well all ATM based links actual= ly) is to start out with the line rate and reduce from there. (The ATM link= layer adjustment effectively reduces to 90% of link rate already as explai= ned above) >=20 >=20 > > But my question is=2C how much is too much? =20 >=20 > So the multistep procedure is roughly as follows. Start with full line r= ate and no link layer adjustments: measure the ping RTT to the nearest host= that responds with no additional traffic=3B that gives you the best case l= atency base line of your link. Next load the link with a speed test and run= the ping again=3B that gives you a bad case (for the worst case you need t= o saturate both up and downlink at the same time=2C while measuring the pin= g RTT). Then activate the link layer adjustments with your line rates speci= fied and repeat the test under load=3B reduce the rates by 5% and repeat. M= ost likely you will notice that each reduction in bandwidth will also reduc= e the latency. You then can see the different possible bandwidth/latency tr= ade-offs possible on your link=2C just pic the one you are most comfortable= with. > Then use your link normally=2C but every now and then=2C when the link i= s loaded repeat the ping test and see whether you are still happy=2C if not= adjust the rates. >=20 >=20 > > The setting of SQM does fix the bufferbloat issue as evidenced by ping = testing and times for packets. With SQM on all packets were 100ms or less.= =20 >=20 > It is interesting to compare this with the ping time to the same host wi= thout any load on your network. >=20 > > With SQM off the times jumped to over 500ms or more during the speed te= sting.=20 > > =20 > > For reference I have set the parameters as indicated in the screenshots= . I have changed only two variables and tested after each change as indica= ted in the grid below. > > =20 > > Que setup script Per packet overhead test results > > Test #1 simple.qos 40 no buffer=2C less throughput <100ms=2C 4.5mbps do= wn > > Test #2 simple.qos 44 no buffer=2C less throughput <100ms=2C 4.5mbps d= own > > Test #3 simplest.qos 40 no buffer=2C less throughput <100ms=2C 4.5mbps = down > > Test #4 simplest.qos 44 no buffer=2C less throughput <100ms=2C 4.5mbps = down >=20 > simple.qos and simplest.qos should have no real effect on either latency= or bandwidth=2C matching your results. The overhead is a tiny bit trickier= =2C if you underestimate the overhead you can=2C under certain conditions= =2C still cause buffer bloat in your modem (but these conditions are tricky= to recreate=2C so underestimation will stochastically show up as latency s= pies every now and then under load)=2C if you overestimate the overhead you= sacrifice more bandwidth than necessary (since the overhead is per packet= =2C this bandwidth sacrifice depends on the typical size of packets you sen= d). The idea is to just pick the rift values for your encapsulation and be = done with. > My current understanding is that 48 bytes is the worst case overhead=2C = but I have not yet seen a link with that=2C 44 bytes however are not uncomm= on=2C so 44 is not the worst place to start out with. (Note it is possible = to empirically measure the overhead on your link=2C and I would be happy to= help you with this=2C just contact me if interested). >=20 > > =20 > > I also read your statement-=20 > > >>>"The CeroWrt development team has been working to nail down a no-bra= iner set of instructions for eliminating bufferbloat - the lag/latency that= kills voice & video chat=2C gaming=2C and overall network responsiveness. = The hard part is that optimal configuration of the Smart Queue Management (= SQM) link is difficult - there are tons of options an ISP can set. Although= CeroWrt can adapt to any of them=2C it's difficult to find out the exact c= haracteristics of the link you have." > > =20 > > What info do I need to get from my ISP to best optimize my connection? >=20 > The actual line rates for down- and uplink as well as the full encapsula= tion information: DHCP or PPPoE or PPPoA=3B VC-MUX or LLC/SNAP=3B VLAN or n= o vlan. >=20 > > =20 > > I also recognize that this could be an issue that requires multiple cha= nges at once. =20 >=20 > You are doing fine=2C all it needs is a few experiments/measurements to = find the best values after that you can basically stick to those. >=20 > > I am curious to know from the experts what your thoughts are on this. = =20 >=20 > That would be Dave then (all I know about is the atm issues=2C as I stil= l have an adel link) >=20 > Best Regards > Sebastian >=20 > > Many thanks in advance! > > =20 > > -Jeremy > > _______________________________________________ > > Cerowrt-users mailing list > > Cerowrt-users@lists.bufferbloat.net > > https://lists.bufferbloat.net/listinfo/cerowrt-users >=20 >=20 >=20 > ------------------------------ >=20 > _______________________________________________ > Cerowrt-users mailing list > Cerowrt-users@lists.bufferbloat.net > https://lists.bufferbloat.net/listinfo/cerowrt-users >=20 >=20 > End of Cerowrt-users Digest=2C Vol 13=2C Issue 2 > ******************************************** =0A= =0A= = --_988dd7de-9b84-4693-a08f-04cd29912e0f_ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
=0A= =0A= =0A=
=0A= =0A= =0A=
Sebastian and Dave=2C =3B thanks to both of you for yo= ur email. =3B I did some more checking on settings in my modem. =3B= The max line rate is listed as 7616kbps down and 512 kbps up in the modem = setup page. =3B My ISP labels this as 6 MB service. =3B The actual = throughput I was getting is 5700kbps and 450 kbps before SQM. =3B
<= br>>=3B>=3B>=3BSo you specified5.7 and 0.45?
No=2C at the time I s= pecified 5130 and 400 based on actual download speeds. =3B

I di= d some more testing this evening. =3B Based on Dave's comment I upgrade= d to the latest build 3.10.26. =3B I also adjusted my ingress and egres= s speeds to 6854 and 460. =3B When I did this my speeds jumped back up = to near normal. =3B So..... I think if I understand correctly you shoul= d set your speed to approximately 90% of the listed line rate in your modem= .

>=3B >=3B What info do I need to get from my ISP to best optim= ize my connection?
>=3B
>=3B=0A= The actual line rates for down- and uplink as well as the full =0A= encapsulation information: DHCP or PPPoE or PPPoA=3B >=3BVC-MUX or LLC/SN= AP=3B =0A= VLAN or no vlan.

My ISP does PPPoE (mine is done on the router inste= ad of the modem - I'm in bridged mode) =3B Encapsulation is LLC. = =3B I also found out the interleave is set at 64. =3B As I understand i= t=2C that has an affect on latency. =3B Anyhow=2C =3B I think that = should cover the most critical info needed. =3B I currently have my per= packet overhead set at 44. =3B
I'll have to dig in and read some m= ore so I better understand what this setting does. =3B I did take a bri= ef look at the RRUL tool. =3B I'll be teaching myself how to set it up = and use it so I can run some more thorough tests this weekend.

>= =3B>=3B(Note it is possible to empirically measure the overhead on your l= ink=2C =0A= and I would be happy to help you with this=2C just contact me if =0A= interested).

Yes=2C I'd be interested in learning more!

I thi= nk both of your emails have helped to clarify what my settings need to be a= nd more importantly I am better understanding HOW you arrived at the approp= riate values and the interplay between them. =3B If I can help with doc= umentation of the SQM setup page on the wiki please let me know.



>=3B From: cerowrt-users-request@lists.bufferbloat.net
&= gt=3B Subject: Cerowrt-users Digest=2C Vol 13=2C Issue 2
>=3B To: cero= wrt-users@lists.bufferbloat.net
>=3B Date: Thu=2C 30 Jan 2014 12:00:01= -0800
>=3B
>=3B Send Cerowrt-users mailing list submissions to<= br>>=3B cerowrt-users@lists.bufferbloat.net
>=3B
>=3B To subs= cribe or unsubscribe via the World Wide Web=2C visit
>=3B https://lis= ts.bufferbloat.net/listinfo/cerowrt-users
>=3B or=2C via email=2C send= a message with subject or body 'help' to
>=3B cerowrt-users-request@= lists.bufferbloat.net
>=3B
>=3B You can reach the person managin= g the list at
>=3B cerowrt-users-owner@lists.bufferbloat.net
>= =3B
>=3B When replying=2C please edit your Subject line so it is more= specific
>=3B than "Re: Contents of Cerowrt-users digest..."
>= =3B
>=3B
>=3B Today's Topics:
>=3B
>=3B 1. Re: SQ= M Setup and Performance (Sebastian Moeller)
>=3B
>=3B
>=3B= ----------------------------------------------------------------------
= >=3B
>=3B Message: 1
>=3B Date: Thu=2C 30 Jan 2014 20:29:50 +0= 100
>=3B From: Sebastian Moeller <=3Bmoeller0@gmx.de>=3B
>=3B= To: Jeremy Tourville <=3Borgan_dr@hotmail.com>=3B
>=3B Cc: "cerow= rt-users@lists.bufferbloat.net"
>=3B <=3Bcerowrt-users@lists.buffer= bloat.net>=3B
>=3B Subject: Re: [Cerowrt-users] SQM Setup and Perfor= mance
>=3B Message-ID: <=3B646BCC1A-0C7A-46AB-827D-717AEA0A2174@gmx.= de>=3B
>=3B Content-Type: text/plain=3B charset=3Diso-8859-1
>= =3B
>=3B Hi Jeremy=2C
>=3B
>=3B On Jan 30=2C 2014=2C at 18= :06 =2C Jeremy Tourville <=3Borgan_dr@hotmail.com>=3B wrote:
>=3B =
>=3B >=3B Hello=2C I followed your excellent instructions here ->=3B >=3B http://www.bufferbloat.net/projects/cerowrt/wiki/Setting_up_= SQM_for_CeroWrt_310
>=3B >=3B
>=3B >=3B I am using build 3.= 10.24-8
>=3B >=3B
>=3B >=3B I am using a DSL line rated at = 6 Mbps down and 512kbps up. My real throughput without SQM enabled is 5.7M= bps down and 450kbps up.
>=3B
>=3B This looks interesting=2C th= e ATM 48 bytes in 53 byte cell encapsulation used in ADSL only leaves 100*4= 8/53 =3D 90.57% percent of the specified line rate available for your traff= ic (but that still contains further per packet overhead). So I would assume= that the actual liberate is >=3B6.3Mbps? Do you have any chance of verif= ying the line rate to the DSLAM? Many modems/dsl-routers have offer a web p= age that gives some statistics and information like the line rate. You will= need the line rate as precise as possible if you want to minimize the ban= dwidth "sacrifice" needed to keep latencies reasonable (if in doubt err on = the too small side though).
>=3B
>=3B >=3B After enabling SQM = my throughput has dropped to approximately 4.5Mbps down and 350 kbps up. <= br>>=3B
>=3B So you specified5.7 and 0.45? Then the link layer ada= ptation mechanism will try to account for the 48in53 problem and cut down y= our available rates by ~10% so the best you can expect is ~5.1 and 0.4. If = you specified 90% of the measured speed you are already at 5.7 * 0.9 (90% o= f measured) =3D 5.13 * 0.9 (fixed ATM overhead) =3D 4.617 and 450 * 0.9 (90= % of measured) =3D 405 * 0.9 (fixed ATM overhead) =3D 364.5. So you can not= rally expect much more than you got.
>=3B
>=3B
>=3B >= =3B Does this seem like an amount that is expected? (within norms?)
>= =3B >=3B It would seem reasonable that I should expect some performance l= oss at the expense of better bufferbloat management based on setting 85-95%= of actual download/upload speeds. Please correct me if I am wrong. :-)>=3B
>=3B Now=2C my recommendation for ADSL links (well all ATM b= ased links actually) is to start out with the line rate and reduce from the= re. (The ATM link layer adjustment effectively reduces to 90% of link rate = already as explained above)
>=3B
>=3B
>=3B >=3B But my = question is=2C how much is too much?
>=3B
>=3B So the multist= ep procedure is roughly as follows. Start with full line rate and no link l= ayer adjustments: measure the ping RTT to the nearest host that responds wi= th no additional traffic=3B that gives you the best case latency base line = of your link. Next load the link with a speed test and run the ping again= =3B that gives you a bad case (for the worst case you need to saturate both= up and downlink at the same time=2C while measuring the ping RTT). Then ac= tivate the link layer adjustments with your line rates specified and repeat= the test under load=3B reduce the rates by 5% and repeat. Most likely you = will notice that each reduction in bandwidth will also reduce the latency. = You then can see the different possible bandwidth/latency trade-offs possib= le on your link=2C just pic the one you are most comfortable with.
>= =3B Then use your link normally=2C but every now and then=2C when the link= is loaded repeat the ping test and see whether you are still happy=2C if n= ot adjust the rates.
>=3B
>=3B
>=3B >=3B The setting of = SQM does fix the bufferbloat issue as evidenced by ping testing and times f= or packets. With SQM on all packets were 100ms or less.
>=3B
&g= t=3B It is interesting to compare this with the ping time to the same host= without any load on your network.
>=3B
>=3B >=3B With SQM off= the times jumped to over 500ms or more during the speed testing.
>= =3B >=3B
>=3B >=3B For reference I have set the parameters as in= dicated in the screenshots. I have changed only two variables and tested a= fter each change as indicated in the grid below.
>=3B >=3B
>= =3B >=3B Que setup script Per packet overhead test results
>=3B >= =3B Test #1 simple.qos 40 no buffer=2C less throughput <=3B100ms=2C 4.5mb= ps down
>=3B >=3B Test #2 simple.qos 44 no buffer=2C less throughpu= t <=3B100ms=2C 4.5mbps down
>=3B >=3B Test #3 simplest.qos 40 no b= uffer=2C less throughput <=3B100ms=2C 4.5mbps down
>=3B >=3B Test = #4 simplest.qos 44 no buffer=2C less throughput <=3B100ms=2C 4.5mbps down=
>=3B
>=3B simple.qos and simplest.qos should have no real effe= ct on either latency or bandwidth=2C matching your results. The overhead is= a tiny bit trickier=2C if you underestimate the overhead you can=2C under = certain conditions=2C still cause buffer bloat in your modem (but these con= ditions are tricky to recreate=2C so underestimation will stochastically sh= ow up as latency spies every now and then under load)=2C if you overestimat= e the overhead you sacrifice more bandwidth than necessary (since the overh= ead is per packet=2C this bandwidth sacrifice depends on the typical size o= f packets you send). The idea is to just pick the rift values for your enca= psulation and be done with.
>=3B My current understanding is that 48 = bytes is the worst case overhead=2C but I have not yet seen a link with tha= t=2C 44 bytes however are not uncommon=2C so 44 is not the worst place to s= tart out with. (Note it is possible to empirically measure the overhead on = your link=2C and I would be happy to help you with this=2C just contact me = if interested).
>=3B
>=3B >=3B
>=3B >=3B I also read = your statement-
>=3B >=3B >=3B>=3B>=3B"The CeroWrt developmen= t team has been working to nail down a no-brainer set of instructions for e= liminating bufferbloat - the lag/latency that kills voice &=3B video cha= t=2C gaming=2C and overall network responsiveness. The hard part is that op= timal configuration of the Smart Queue Management (SQM) link is difficult -= there are tons of options an ISP can set. Although CeroWrt can adapt to an= y of them=2C it's difficult to find out the exact characteristics of the li= nk you have."
>=3B >=3B
>=3B >=3B What info do I need to ge= t from my ISP to best optimize my connection?
>=3B
>=3B The act= ual line rates for down- and uplink as well as the full encapsulation infor= mation: DHCP or PPPoE or PPPoA=3B VC-MUX or LLC/SNAP=3B VLAN or no vlan.>=3B
>=3B >=3B
>=3B >=3B I also recognize that this cou= ld be an issue that requires multiple changes at once.
>=3B
>= =3B You are doing fine=2C all it needs is a few experiments/measurements t= o find the best values after that you can basically stick to those.
>= =3B
>=3B >=3B I am curious to know from the experts what your thoug= hts are on this.
>=3B
>=3B That would be Dave then (all I kno= w about is the atm issues=2C as I still have an adel link)
>=3B
&g= t=3B Best Regards
>=3B Sebastian
>=3B
>=3B >=3B Many tha= nks in advance!
>=3B >=3B
>=3B >=3B -Jeremy
>=3B >= =3B _______________________________________________
>=3B >=3B Cerowr= t-users mailing list
>=3B >=3B Cerowrt-users@lists.bufferbloat.net>=3B >=3B https://lists.bufferbloat.net/listinfo/cerowrt-users
>= =3B
>=3B
>=3B
>=3B ------------------------------
>= =3B
>=3B _______________________________________________
>=3B Ce= rowrt-users mailing list
>=3B Cerowrt-users@lists.bufferbloat.net
&= gt=3B https://lists.bufferbloat.net/listinfo/cerowrt-users
>=3B
&g= t=3B
>=3B End of Cerowrt-users Digest=2C Vol 13=2C Issue 2
>=3B = ********************************************
=0A=
=0A=
= --_988dd7de-9b84-4693-a08f-04cd29912e0f_--