Date: Mon, 21 Oct 2024 10:42:49 -0400 From: Cheng Cui <cc@freebsd.org> To: void <void@f-m.fm> Cc: "freebsd-net@FreeBSD.org" <freebsd-net@freebsd.org> Subject: Re: Performance test for CUBIC in stable/14 Message-ID: <CAGaXuiKD-b4PGrqfxy9zk-BRxU==HMc9KshqyJGzH8saeOLf1A@mail.gmail.com> In-Reply-To: <ZxJe8e8sRU9NCHv4@vm2> References: <20240912181618.7895d10ad5ff2ebae9883192@gmail.com> <Zw3YGPIBC_s-q-Vg@vm2> <Zw3dlDuaSXsIG-yX@vm2> <CAGaXuiLG4Qv2KgDtJveN3adLh6b8tEzC4j6OYfLo1_dj8BKZWQ@mail.gmail.com> <Zw_6SNSEW5mXMy_I@vm2> <CAGaXuiL_Fk=_FtU23RoJBP0WguU8_-y3Tb1iZ=T2unVv-2aRpA@mail.gmail.com> <ZxDdnSngi13UyZSL@vm2> <CAGaXuiK52ufOs-eVfKHAYpGxRyGfCMpbbZy4GcShybTzHcjDGg@mail.gmail.com> <ZxEqfISBJIW75p_P@vm2> <CAGaXuiJW5QW1eFqjXtSO%2BojOX5G12B56%2BnzE9zaVR7nB33VjWA@mail.gmail.com> <ZxJe8e8sRU9NCHv4@vm2>
next in thread | previous in thread | raw e-mail | index | archive | help
--000000000000db2dbc0624fdac81 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Change the subject to `Performance test for CUBIC in stable/14`, was `Re: Performance issues with vnet jails + epair + bridge`. I actually prepared two patches, one depends on the other: https://reviews.freebsd.org/D47218 << apply this patch firstly https://reviews.freebsd.org/D47213 << apply this patch secondly Please let me know if you have any questions. It passed my local test on a 1% drop rate link, with nearly 3.9X performance improvement. before patching: [ ID] Interval Transfer Bitrate Retr Cwnd [ 5] 0.00-1.07 sec 27.0 MBytes 211 Mbits/sec 207 12.8 KBytes [ 5] 1.07-2.12 sec 18.0 MBytes 144 Mbits/sec 151 11.4 KBytes [ 5] 2.12-3.07 sec 22.0 MBytes 194 Mbits/sec 189 11.8 KBytes [ 5] 3.07-4.07 sec 23.0 MBytes 194 Mbits/sec 203 10.8 KBytes [ 5] 4.07-5.06 sec 18.0 MBytes 153 Mbits/sec 165 15.0 KBytes [ 5] 5.06-6.11 sec 25.0 MBytes 200 Mbits/sec 227 26.9 KBytes [ 5] 6.11-7.07 sec 23.0 MBytes 201 Mbits/sec 189 10.7 KBytes [ 5] 7.07-8.07 sec 23.0 MBytes 193 Mbits/sec 215 12.7 KBytes [ 5] 8.07-9.03 sec 22.0 MBytes 191 Mbits/sec 185 10.4 KBytes [ 5] 9.03-10.09 sec 24.0 MBytes 191 Mbits/sec 219 1.41 KBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bitrate Retr [ 5] 0.00-10.09 sec 225 MBytes 187 Mbits/sec 1950 sender [ 5] 0.00-10.14 sec 225 MBytes 186 Mbits/sec receiver iperf Done. after patching: [ ID] Interval Transfer Bitrate Retr Cwnd [ 5] 0.00-1.08 sec 118 MBytes 919 Mbits/sec 872 47.5 KBytes [ 5] 1.08-2.07 sec 109 MBytes 917 Mbits/sec 849 56.9 KBytes [ 5] 2.07-3.09 sec 111 MBytes 915 Mbits/sec 810 54.5 KBytes [ 5] 3.09-4.07 sec 108 MBytes 922 Mbits/sec 831 43.8 KBytes [ 5] 4.07-5.07 sec 109 MBytes 919 Mbits/sec 813 31.1 KBytes [ 5] 5.07-6.02 sec 103 MBytes 913 Mbits/sec 815 60.2 KBytes [ 5] 6.02-7.07 sec 116 MBytes 921 Mbits/sec 850 58.2 KBytes [ 5] 7.07-8.06 sec 109 MBytes 922 Mbits/sec 822 46.4 KBytes [ 5] 8.06-9.06 sec 109 MBytes 913 Mbits/sec 843 59.4 KBytes [ 5] 9.06-10.07 sec 110 MBytes 917 Mbits/sec 815 42.3 KBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bitrate Retr [ 5] 0.00-10.07 sec 1.08 GBytes 918 Mbits/sec 8320 sender [ 5] 0.00-10.11 sec 1.08 GBytes 915 Mbits/sec receiver iperf Done. cc On Fri, Oct 18, 2024 at 9:13=E2=80=AFAM void <void@f-m.fm> wrote: > On Fri, Oct 18, 2024 at 07:28:49AM -0400, Cheng Cui wrote: > > >The patch is a TCP congestion control algorithm improvement. So to > >be clear, it only impacts a TCP data sender. These hosts are just traffi= c > >forwarders, not TCP sender/receiver. > > > >I can send you a patch for the FreeBSD 14/stable to test performance > >improvement. > > sure! I'm happy to test > > >If you think FreeBSD 14.x releases are critical to your business, or you > >can't wait until FreeBSD 15 is released, you can file a request to add > this > >patch in the 14.x. I think https://bugs.freebsd.org/bugzilla/ is where a > >user can file such request. If it is approved, which I am not sure if > >it can be approved, then maybe it can be included in the next 14.x > release. > > thank you, I'll try :D > -- > > --=20 Best Regards, Cheng Cui --000000000000db2dbc0624fdac81 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr"><div>Change the subject to `Performance test for CUBIC in = stable/14`, was `<span style=3D"font-family:"Google Sans",Roboto,= RobotoDraft,Helvetica,Arial,sans-serif;font-size:14.000001px">Re: Performan= ce issues with vnet jails + epair + bridge</span>`.</div><div><br></div><di= v>I actually prepared two patches, one depends on the other:</div><div><br>= </div><div><font face=3D"monospace"><a href=3D"https://reviews.freebsd.org/= D47218">https://reviews.freebsd.org/D47218</a>; =C2=A0 =C2=A0 << apply= this patch firstly<br></font></div><div><font face=3D"monospace"><a href= =3D"https://reviews.freebsd.org/D47213">https://reviews.freebsd.org/D47213<= /a> =C2=A0 =C2=A0 << apply this patch secondly</font><br></div><div><= br></div><div>Please let me know if you have=C2=A0any questions.</div><div>= <br></div><div>It passed my local test on a 1% drop rate link, with nearly = 3.9X performance improvement.</div><div><br></div><div>before patching:</di= v><div>[ ID] Interval =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 Transfer =C2=A0 = =C2=A0 Bitrate =C2=A0 =C2=A0 =C2=A0 =C2=A0 Retr =C2=A0Cwnd<br>[ =C2=A05] = =C2=A0 0.00-1.07 =C2=A0 sec =C2=A027.0 MBytes =C2=A0 211 Mbits/sec =C2=A020= 7 =C2=A0 12.8 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 1.07-2.12 = =C2=A0 sec =C2=A018.0 MBytes =C2=A0 144 Mbits/sec =C2=A0151 =C2=A0 11.4 KBy= tes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 2.12-3.07 =C2=A0 sec =C2=A02= 2.0 MBytes =C2=A0 194 Mbits/sec =C2=A0189 =C2=A0 11.8 KBytes =C2=A0 =C2=A0 = =C2=A0 <br>[ =C2=A05] =C2=A0 3.07-4.07 =C2=A0 sec =C2=A023.0 MBytes =C2=A0 = 194 Mbits/sec =C2=A0203 =C2=A0 10.8 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2= =A05] =C2=A0 4.07-5.06 =C2=A0 sec =C2=A018.0 MBytes =C2=A0 153 Mbits/sec = =C2=A0165 =C2=A0 15.0 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 5.0= 6-6.11 =C2=A0 sec =C2=A025.0 MBytes =C2=A0 200 Mbits/sec =C2=A0227 =C2=A0 2= 6.9 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 6.11-7.07 =C2=A0 sec = =C2=A023.0 MBytes =C2=A0 201 Mbits/sec =C2=A0189 =C2=A0 10.7 KBytes =C2=A0 = =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 7.07-8.07 =C2=A0 sec =C2=A023.0 MBytes = =C2=A0 193 Mbits/sec =C2=A0215 =C2=A0 12.7 KBytes =C2=A0 =C2=A0 =C2=A0 <br>= [ =C2=A05] =C2=A0 8.07-9.03 =C2=A0 sec =C2=A022.0 MBytes =C2=A0 191 Mbits/s= ec =C2=A0185 =C2=A0 10.4 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 = 9.03-10.09 =C2=A0sec =C2=A024.0 MBytes =C2=A0 191 Mbits/sec =C2=A0219 =C2= =A0 1.41 KBytes =C2=A0 =C2=A0 =C2=A0 <br>- - - - - - - - - - - - - - - - - = - - - - - - - -<br>[ ID] Interval =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 Transf= er =C2=A0 =C2=A0 Bitrate =C2=A0 =C2=A0 =C2=A0 =C2=A0 Retr<br>[ =C2=A05] =C2= =A0 0.00-10.09 =C2=A0sec =C2=A0 225 MBytes =C2=A0 187 Mbits/sec =C2=A01950 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 sender<br>[ =C2=A05] =C2=A0 0.00-= 10.14 =C2=A0sec =C2=A0 225 MBytes =C2=A0 186 Mbits/sec =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0receiver<br><br>iperf Done.<br></= div><div><br></div><div>after patching:</div><div>[ ID] Interval =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 Transfer =C2=A0 =C2=A0 Bitrate =C2=A0 =C2=A0 =C2= =A0 =C2=A0 Retr =C2=A0Cwnd<br>[ =C2=A05] =C2=A0 0.00-1.08 =C2=A0 sec =C2=A0= 118 MBytes =C2=A0 919 Mbits/sec =C2=A0872 =C2=A0 47.5 KBytes =C2=A0 =C2=A0= =C2=A0 <br>[ =C2=A05] =C2=A0 1.08-2.07 =C2=A0 sec =C2=A0 109 MBytes =C2=A0= 917 Mbits/sec =C2=A0849 =C2=A0 56.9 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2= =A05] =C2=A0 2.07-3.09 =C2=A0 sec =C2=A0 111 MBytes =C2=A0 915 Mbits/sec = =C2=A0810 =C2=A0 54.5 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 3.0= 9-4.07 =C2=A0 sec =C2=A0 108 MBytes =C2=A0 922 Mbits/sec =C2=A0831 =C2=A0 4= 3.8 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 4.07-5.07 =C2=A0 sec = =C2=A0 109 MBytes =C2=A0 919 Mbits/sec =C2=A0813 =C2=A0 31.1 KBytes =C2=A0 = =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 5.07-6.02 =C2=A0 sec =C2=A0 103 MBytes = =C2=A0 913 Mbits/sec =C2=A0815 =C2=A0 60.2 KBytes =C2=A0 =C2=A0 =C2=A0 <br>= [ =C2=A05] =C2=A0 6.02-7.07 =C2=A0 sec =C2=A0 116 MBytes =C2=A0 921 Mbits/s= ec =C2=A0850 =C2=A0 58.2 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 = 7.07-8.06 =C2=A0 sec =C2=A0 109 MBytes =C2=A0 922 Mbits/sec =C2=A0822 =C2= =A0 46.4 KBytes =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 8.06-9.06 =C2=A0= sec =C2=A0 109 MBytes =C2=A0 913 Mbits/sec =C2=A0843 =C2=A0 59.4 KBytes = =C2=A0 =C2=A0 =C2=A0 <br>[ =C2=A05] =C2=A0 9.06-10.07 =C2=A0sec =C2=A0 110 = MBytes =C2=A0 917 Mbits/sec =C2=A0815 =C2=A0 42.3 KBytes =C2=A0 =C2=A0 =C2= =A0 <br>- - - - - - - - - - - - - - - - - - - - - - - - -<br>[ ID] Interval= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 Transfer =C2=A0 =C2=A0 Bitrate =C2=A0 = =C2=A0 =C2=A0 =C2=A0 Retr<br>[ =C2=A05] =C2=A0 0.00-10.07 =C2=A0sec =C2=A01= .08 GBytes =C2=A0 918 Mbits/sec =C2=A08320 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 sender<br>[ =C2=A05] =C2=A0 0.00-10.11 =C2=A0sec =C2=A01.08 GByt= es =C2=A0 915 Mbits/sec =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0receiver<br><br>iperf Done.<br></div><div><br></div>cc<div><br= ><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Fri, O= ct 18, 2024 at 9:13=E2=80=AFAM void <<a href=3D"mailto:void@f-m.fm">void= @f-m.fm</a>> wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"= margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-style:solid;bord= er-left-color:rgb(204,204,204);padding-left:1ex">On Fri, Oct 18, 2024 at 07= :28:49AM -0400, Cheng Cui wrote:<br> <br> >The patch is a TCP congestion control algorithm improvement. So to<br> >be clear, it only impacts a TCP data sender. These hosts are just traff= ic<br> >forwarders, not TCP sender/receiver.<br> ><br> >I can send you a patch for the FreeBSD 14/stable to test performance<br= > >improvement.<br> <br> sure! I'm happy to test<br> <br> >If you think FreeBSD 14.x releases are critical to your business, or yo= u<br> >can't wait until FreeBSD 15 is released, you can file a request to = add this <br> >patch in the 14.x. I think <a href=3D"https://bugs.freebsd.org/bugzilla= /" rel=3D"noreferrer" target=3D"_blank">https://bugs.freebsd.org/bugzilla/<= /a> is where a <br> >user can file such request. If it is approved, which I am not sure if <= br> >it can be approved, then maybe it can be included in the next 14.x rele= ase.<br> <br> thank you, I'll try :D<br> -- <br> <br> </blockquote></div><br clear=3D"all"><div><br></div><span class=3D"gmail_si= gnature_prefix">-- </span><br><div dir=3D"ltr" class=3D"gmail_signature"><d= iv dir=3D"ltr"><div></div>Best Regards,<div>Cheng Cui</div></div></div></di= v></div> --000000000000db2dbc0624fdac81--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAGaXuiKD-b4PGrqfxy9zk-BRxU==HMc9KshqyJGzH8saeOLf1A>