Date: Sat, 23 Dec 2023 14:19:28 -0700 From: Warner Losh <imp@bsdimp.com> To: void <void@f-m.fm> Cc: freebsd-fs <freebsd-fs@freebsd.org> Subject: Re: measuring swap partition speed Message-ID: <CANCZdfpbuOo4qpjifcmq8i8Kz_7a4DWcVNkqJbMEwnkLA-W6Lw@mail.gmail.com> In-Reply-To: <2dcc21bb-f446-41fa-871b-2c37cdf9afe0@app.fastmail.com> References: <A6B49CB2-2EEC-4F1F-8777-D7EF65C71E5E.ref@yahoo.com> <A6B49CB2-2EEC-4F1F-8777-D7EF65C71E5E@yahoo.com> <2dcc21bb-f446-41fa-871b-2c37cdf9afe0@app.fastmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--000000000000fbc1c2060d33e351 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Thu, Dec 21, 2023 at 5:20=E2=80=AFPM void <void@f-m.fm> wrote: > Hi Mark & Warner, > > I tried using fio with Warner's suggestions as a template, and > pasted the results in the latter part of this message. > > On Thu, 21 Dec 2023, at 21:03, Mark Millard wrote: > > > # sysctl vm.stats.vm.v_page_size > > vm.stats.vm.v_page_size: 4096 > > Here, this setting is: > # sysctl vm.stats.vm.v_page_size > vm.stats.vm.v_page_size: 4096 > > > # sysctl vm.phys_pager_cluster > > vm.phys_pager_cluster: 1024 > > It is 1024 here, too. > > Only the one usb3 port is occupied. > > fio output follows. No idea yet if the settings used are suitable for thi= s > context > and would welcome suggestions. I think --filename can be a device? Output > suggests > it used hw.ncpu instead of --numjobs=3D8 > > ~~~~~~~ > # fio --name=3Drandread --ioengine=3Dposixaio --rw=3Drandread --direct=3D= 1 --bs=3D8k > --refill_buffers --norandommap --randrepeat=3D0 --iodepth=3D32 --numjobs= =3D8 > --runtime=3D60 --group_reporting --thread --size=3D2048M > randread: (g=3D0): rw=3Drandread, bs=3D(R) 8192B-8192B, (W) 8192B-8192B, = (T) > 8192B-8192B, ioengine=3Dposixaio, iodepth=3D32 > .. > fio-3.36 > Starting 8 threads > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > randread: Laying out IO file (1 file / 2048MiB) > Jobs: 4 (f=3D4): [_(1),r(2),_(2),r(1),_(1),r(1)][21.2%][r=3D288KiB/s][r= =3D36 > IOPS][eta 04m:05s] > randread: (groupid=3D0, jobs=3D8): err=3D 0: pid=3D135125: Thu Dec 21 16:= 43:00 2023 > read: IOPS=3D43, BW=3D351KiB/s (359kB/s)(22.6MiB/65986msec) > slat (nsec): min=3D889, max=3D1948.6k, avg=3D4172.45, stdev=3D36488.0= 7 > clat (msec): min=3D1108, max=3D11660, avg=3D5644.23, stdev=3D1282.12 > lat (msec): min=3D1108, max=3D11660, avg=3D5644.24, stdev=3D1282.12 > clat percentiles (msec): > | 1.00th=3D[ 1183], 5.00th=3D[ 3171], 10.00th=3D[ 4933], 20.00th= =3D[ 5269], > | 30.00th=3D[ 5470], 40.00th=3D[ 5604], 50.00th=3D[ 5738], 60.00th= =3D[ 5873], > | 70.00th=3D[ 5940], 80.00th=3D[ 6074], 90.00th=3D[ 6342], 95.00th= =3D[ 6812], > | 99.00th=3D[10671], 99.50th=3D[10939], 99.90th=3D[11610], 99.95th= =3D[11610], > | 99.99th=3D[11610] > bw ( KiB/s): min=3D 208, max=3D 3760, per=3D100.00%, avg=3D1535.05, > stdev=3D128.23, samples=3D245 > iops : min=3D 26, max=3D 470, avg=3D191.88, stdev=3D16.03, s= amples=3D245 > lat (msec) : 2000=3D3.25%, >=3D2000=3D96.75% > cpu : usr=3D0.00%, sys=3D0.12%, ctx=3D22712, majf=3D0, minf=3D= 0 > IO depths : 1=3D0.3%, 2=3D0.6%, 4=3D1.1%, 8=3D4.9%, 16=3D69.6%, 32= =3D23.6%, > >=3D64=3D0.0% > submit : 0=3D0.0%, 4=3D100.0%, 8=3D0.0%, 16=3D0.0%, 32=3D0.0%, 64= =3D0.0%, > >=3D64=3D0.0% > complete : 0=3D0.0%, 4=3D94.7%, 8=3D2.9%, 16=3D1.5%, 32=3D0.9%, 64= =3D0.0%, > >=3D64=3D0.0% > issued rwts: total=3D2895,0,0,0 short=3D0,0,0,0 dropped=3D0,0,0,0 > latency : target=3D0, window=3D0, percentile=3D100.00%, depth=3D32 > > Run status group 0 (all jobs): > READ: bw=3D351KiB/s (359kB/s), 351KiB/s-351KiB/s (359kB/s-359kB/s), > io=3D22.6MiB (23.7MB), run=3D65986-65986msec > # > 5s+ of latency on the average, max latency of 12s! Woof. No wonder you hate life. Warner --000000000000fbc1c2060d33e351 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">= <div dir=3D"ltr" class=3D"gmail_attr">On Thu, Dec 21, 2023 at 5:20=E2=80=AF= PM 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;bor= der-left:1px solid rgb(204,204,204);padding-left:1ex">Hi Mark & Warner,= <br> <br> I tried using fio with Warner's suggestions as a template, and<br> pasted the results in the latter part of this message.<br> <br> On Thu, 21 Dec 2023, at 21:03, Mark Millard wrote:<br> <br> > # sysctl vm.stats.vm.v_page_size<br> > vm.stats.vm.v_page_size: 4096<br> <br> Here, this setting is:<br> # sysctl vm.stats.vm.v_page_size<br> vm.stats.vm.v_page_size: 4096<br> <br> > # sysctl vm.phys_pager_cluster<br> > vm.phys_pager_cluster: 1024<br> <br> It is 1024 here, too.<br> <br> Only the one usb3 port is occupied.<br> <br> fio output follows. No idea yet if the settings used are suitable for this = context<br> and would welcome suggestions. I think --filename can be a device? Output s= uggests<br> it used hw.ncpu instead of --numjobs=3D8<br> <br> ~~~~~~~<br> # fio --name=3Drandread --ioengine=3Dposixaio --rw=3Drandread --direct=3D1 = --bs=3D8k --refill_buffers --norandommap --randrepeat=3D0 --iodepth=3D32 --= numjobs=3D8 --runtime=3D60 --group_reporting --thread --size=3D2048M<br> randread: (g=3D0): rw=3Drandread, bs=3D(R) 8192B-8192B, (W) 8192B-8192B, (T= ) 8192B-8192B, ioengine=3Dposixaio, iodepth=3D32<br> ..<br> fio-3.36<br> Starting 8 threads<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> randread: Laying out IO file (1 file / 2048MiB)<br> Jobs: 4 (f=3D4): [_(1),r(2),_(2),r(1),_(1),r(1)][21.2%][r=3D288KiB/s][r=3D3= 6 IOPS][eta 04m:05s]<br> randread: (groupid=3D0, jobs=3D8): err=3D 0: pid=3D135125: Thu Dec 21 16:43= :00 2023<br> =C2=A0 read: IOPS=3D43, BW=3D351KiB/s (359kB/s)(22.6MiB/65986msec)<br> =C2=A0 =C2=A0 slat (nsec): min=3D889, max=3D1948.6k, avg=3D4172.45, stdev= =3D36488.07<br> =C2=A0 =C2=A0 clat (msec): min=3D1108, max=3D11660, avg=3D5644.23, stdev=3D= 1282.12<br> =C2=A0 =C2=A0 =C2=A0lat (msec): min=3D1108, max=3D11660, avg=3D5644.24, std= ev=3D1282.12<br> =C2=A0 =C2=A0 clat percentiles (msec):<br> =C2=A0 =C2=A0 =C2=A0|=C2=A0 1.00th=3D[ 1183],=C2=A0 5.00th=3D[ 3171], 10.00= th=3D[ 4933], 20.00th=3D[ 5269],<br> =C2=A0 =C2=A0 =C2=A0| 30.00th=3D[ 5470], 40.00th=3D[ 5604], 50.00th=3D[ 573= 8], 60.00th=3D[ 5873],<br> =C2=A0 =C2=A0 =C2=A0| 70.00th=3D[ 5940], 80.00th=3D[ 6074], 90.00th=3D[ 634= 2], 95.00th=3D[ 6812],<br> =C2=A0 =C2=A0 =C2=A0| 99.00th=3D[10671], 99.50th=3D[10939], 99.90th=3D[1161= 0], 99.95th=3D[11610],<br> =C2=A0 =C2=A0 =C2=A0| 99.99th=3D[11610]<br> =C2=A0 =C2=A0bw (=C2=A0 KiB/s): min=3D=C2=A0 208, max=3D 3760, per=3D100.00= %, avg=3D1535.05, stdev=3D128.23, samples=3D245<br> =C2=A0 =C2=A0iops=C2=A0 =C2=A0 =C2=A0 =C2=A0 : min=3D=C2=A0 =C2=A026, max= =3D=C2=A0 470, avg=3D191.88, stdev=3D16.03, samples=3D245<br> =C2=A0 lat (msec)=C2=A0 =C2=A0: 2000=3D3.25%, >=3D2000=3D96.75%<br> =C2=A0 cpu=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 : usr=3D0.00%, sys=3D0.12%, ct= x=3D22712, majf=3D0, minf=3D0<br> =C2=A0 IO depths=C2=A0 =C2=A0 : 1=3D0.3%, 2=3D0.6%, 4=3D1.1%, 8=3D4.9%, 16= =3D69.6%, 32=3D23.6%, >=3D64=3D0.0%<br> =C2=A0 =C2=A0 =C2=A0submit=C2=A0 =C2=A0 : 0=3D0.0%, 4=3D100.0%, 8=3D0.0%, 1= 6=3D0.0%, 32=3D0.0%, 64=3D0.0%, >=3D64=3D0.0%<br> =C2=A0 =C2=A0 =C2=A0complete=C2=A0 : 0=3D0.0%, 4=3D94.7%, 8=3D2.9%, 16=3D1.= 5%, 32=3D0.9%, 64=3D0.0%, >=3D64=3D0.0%<br> =C2=A0 =C2=A0 =C2=A0issued rwts: total=3D2895,0,0,0 short=3D0,0,0,0 dropped= =3D0,0,0,0<br> =C2=A0 =C2=A0 =C2=A0latency=C2=A0 =C2=A0: target=3D0, window=3D0, percentil= e=3D100.00%, depth=3D32<br> <br> Run status group 0 (all jobs):<br> =C2=A0 =C2=A0READ: bw=3D351KiB/s (359kB/s), 351KiB/s-351KiB/s (359kB/s-359k= B/s), io=3D22.6MiB (23.7MB), run=3D65986-65986msec<br> #<br></blockquote><div><br></div><div>5s+ of latency on the average, max la= tency of 12s!=C2=A0 Woof. No wonder you hate life.</div><div><br></div><div= >Warner</div></div></div> --000000000000fbc1c2060d33e351--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfpbuOo4qpjifcmq8i8Kz_7a4DWcVNkqJbMEwnkLA-W6Lw>