Hi,
I'm trying to debug some general network slowness with my 6.8 server
(i7-3930k) that seems to affect all protocols (e.g. Samba capping at
70MB/s, FTP at 45MB/s for upload). I've run some iperf3/tcpbench tests
and the results seems low even when running both client and server
on the same machine to eliminate the actual network.
Setting the client to localhost, I get an average of about 10Gbit/s in
iperf3.
[ 5] 0.00-10.00 sec 11.3 GBytes 9.74 Gbits/sec sender
[ 5] 0.00-10.00 sec 11.3 GBytes 9.74 Gbits/sec receiver
tcpbench is a bit quicker:
Conn: 1 Mbps:11585.269 Peak Mbps:11602.117 Avg Mbps:11585.269
Conn: 1 Mbps:11580.000 Peak Mbps:11602.117 Avg Mbps:11580.000
Conn: 1 Mbps:11583.638 Peak Mbps:11602.117 Avg Mbps:11583.638
Conn: 1 Mbps:10029.172 Peak Mbps:11602.117 Avg Mbps:10029.172
Running with the IP address of the machine:
[ 5] 0.00-10.00 sec 1.51 GBytes 1.30 Gbits/sec sender
[ 5] 0.00-10.00 sec 1.51 GBytes 1.30 Gbits/sec receiver
Conn: 1 Mbps: 1026.018 Peak Mbps: 1543.406 Avg Mbps: 1026.018
Conn: 1 Mbps: 977.226 Peak Mbps: 1543.406 Avg Mbps: 977.226
Conn: 1 Mbps: 1281.376 Peak Mbps: 1543.406 Avg Mbps: 1281.376
Conn: 1 Mbps: 1128.490 Peak Mbps: 1543.406 Avg Mbps: 1128.490
smbclient to IP address is around 100-120MB/s depending on Samba
settings. To localhost it goes up to 200MB/s. This is copying between
SSDs which manage 500MB/s directly.
Those tests were with PF disabled. PF enabled is basically the same:
[ 5] 0.00-10.00 sec 1.49 GBytes 1.28 Gbits/sec sender
[ 5] 0.00-10.00 sec 1.49 GBytes 1.28 Gbits/sec receiver
With UDP, iperf3 seems broken (1Mbit/s), with tcpbench:
localhost:
Elapsed: 2 Mbps: 912.534 Peak Mbps: 926.653 Tx PPS: 77491
Elapsed: 21000 Mbps: 861.850 Peak Mbps: 926.653 Tx PPS: 73187
local IP:
Elapsed: 16000 Mbps: 908.789 Peak Mbps: 917.009 Tx PPS: 77173
Elapsed: 17000 Mbps: 788.132 Peak Mbps: 917.009 Tx PPS: 66927
systate with tcpbench running with TCP & IP address:
3 users Load 1.31 0.64 0.32 caleb.home.duncanma 14:29:14
memory totals (in KB)PAGING SWAPPING Interrupts
real virtual free in out in out 204 total
Active 211624211624 11526980 ops200 clock
All 4660156 4660156 19920940 pages3 ipi
1 em0
Proc:r d s wCsw Trp Sys Int Sof Flt forks ehci0
1 105 299265634 1171 fkppw azalia1
fksvm ahci0
0.0%Int 0.7%Spn 17.5%Sys 0.8%Usr 81.0%Idle pwait ehci1
||||||||||| relck ahci1
= rlkok pckbc0
noram
Namei Sys-cacheProc-cacheNo-cache ndcpy
Calls hits%hits %miss % fltcp
zfod
cow
Disks sd0 sd1 sd2 sd3 134892 fmin
seeks 179856 ftarg
xfers itarg
speed3K 3K 2 wired
sec 0.0 0.0 pdfre
pdscn
pzidl 211289 IPKTS
13 kmape 211288 OPKTS
and top:
load averages: 1.39, 1.01, 0.54
66 processes: 63 idle, 3 on processor
CPU00 states: 0.0% user, 0.0% nice, 0.0% sys, 0.0% spin, 0.0% intr, 100%
idle
CPU01 states: 0.0% user, 0.0% nice, 0.0% sys, 0.0% spin, 0.0% intr, 100%
idle
CPU02 states: 1.0% user, 0.0% nice, 53.3% sys, 2.0% spin, 0.0% intr, 43.7%
idle
CPU03 states: 0.4% user, 0.0% nice, 48.9% sys, 2.0% spin, 0.0% intr, 48.7%
idle
CPU04 states: 1.2% user, 0.0% nice, 7.8% sys, 0.0% spin, 0.0% intr, 91.0%
idle
CPU05 states: 0.0% user, 0.0% nice, 0.0% sys, 0.0% spin, 0.0% intr, 100%
idle
Memory: Real: 206M/4550M act/tot Free: 11G Cache: 3781M Swap: 0K/8197M
PID USERNAME PRI NICE SIZE RES STATE WAIT TIMECPU COMMAND
55082 duncan400 680K 2060K onproc/2 - 1:17 13.96% tcpbench
41980 duncan330 660K 2040K onproc/4 - 0:05 6.30% tcpbench
I also tried a different machine (i9-10850K)