Date: Sun, 27 Oct 2024 14:54:22 +0000 From: void <void@f-m.fm> To: freebsd-net@freebsd.org Subject: Re: Performance test for CUBIC in stable/14 Message-ID: <4a5dd205-6c8f-4964-b254-7092aa590602@app.fastmail.com> In-Reply-To: <CAGaXuiKaUurcPwwB-gJYMU2ce3i3cn7mT0JFjZCkM_dU5gBNXg@mail.gmail.com> References: <ZxJe8e8sRU9NCHv4@vm2> <CAGaXuiKD-b4PGrqfxy9zk-BRxU==HMc9KshqyJGzH8saeOLf1A@mail.gmail.com> <ZxaccxFblDt0UQWR@int21h> <CAGaXui%2BQ7wCM1dAKis%2BvNaNJ5uODeiC765hoXWT4OBtT7npprw@mail.gmail.com> <ZxfFRg3tYtdQt0hM@vm2> <CAGaXuiJvC2i5yxEaaDmHMoadPzkk3oqQOzg1yiqBuhTR%2B=R9Sg@mail.gmail.com> <ZxgHOlRaCR6Joqdv@vm2> <CAGaXui%2BEYmRhOdwOqFRbVsboCkrrWmnHnWRSqsSAgzbn5ug6bg@mail.gmail.com> <ZxkQiHWlDz28az-N@vm2> <CAGaXuiLuuXW_gFMq=--1Z2rKZ4ZZUiV52BnW10FxDYZc6vBZ-Q@mail.gmail.com> <Zxlt-dQHwz_Gl_Sz@int21h> <CAGaXuiKaUurcPwwB-gJYMU2ce3i3cn7mT0JFjZCkM_dU5gBNXg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 25 Oct 2024, at 13:13, Cheng Cui wrote: > Here is my example. I am using two 6-core/12-threads desktops for my > bhyve servers. > CPU: AMD Ryzen 5 5560U with Radeon Graphics (2295.75-MHz > K8-class CPU) > > You can find test results on VMs from my wiki: > https://wiki.freebsd.org/chengcui/testD46046 > > All the CPU utilization results are low, especially for these > throughput over 900 Mb/s. > > cc I have to tabulate my results better :D Those are some p impressive performance gains for your modified cc_cubic against % packet loss. This is a freebsd14-stable vm with your patches applied. iperf3 -B 192.168.1.13 -c 192.168.1.232 -l 1M -t 60 -i 1 -VC cubic Test Complete. Summary Results: [ ID] Interval Transfer Bitrate Retr [ 5] 0.00-60.07 sec 5.92 GBytes 847 Mbits/sec 1 sender [ 5] 0.00-60.09 sec 5.92 GBytes 846 Mbits/sec receiver CPU Utilization: local/sender 90.6% (0.0%u/90.6%s), remote/receiver 23.8% (2.0%u/21.8%s) snd_tcp_congestion cubic rcv_tcp_congestion cubic here's the Alpine linux vm: vm7-alpine:~$ iperf3 -B 192.168.1.65 -c 192.168.1.232 -l 1M -t 60 -i 1 -VC cubic Test Complete. Summary Results: [ ID] Interval Transfer Bitrate Retr [ 5] 0.00-60.02 sec 6.58 GBytes 941 Mbits/sec 24 sender [ 5] 0.00-60.03 sec 6.58 GBytes 941 Mbits/sec receiver CPU Utilization: local/sender 8.2% (0.1%u/8.2%s), remote/receiver 26.7% (2.1%u/24.5%s) snd_tcp_congestion cubic rcv_tcp_congestion cubic Load on the bhyve host was between 3.00 & 4.00 during the tests. It was running a poudriere build (PARALLEL_JOBS=1) Why the massive difference in load on the vms? I'm trying to look at something that's maybe related, but not in a congested environment context. I think it's deeper than that. I need to try 'options RSS' in the freebsd guests as well, (see https://lists.freebsd.org/archives/freebsd-net/2024-September/005531.html ) --
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4a5dd205-6c8f-4964-b254-7092aa590602>