|
服务端windows server 2022，博通57504 25G 网卡，4 NVME 4.0 做存储池。客户端windows server 2022 connectx4 25G网卡。
客户端通过SMB访问服务端存储池，两路25G 线路网卡直连，测试在win下使用FIO，4k随机读写，队列32，Dierect IO，20线程。
命令fio --rw=randwrite --bs=4k --iodepth=32 --direct=1 --filename=Y\:\\fio_system_disk_test --name=fiotest --ioengine=windowsaio --thread --group_reporting --numjobs=20 --size=1G --runtime=50 --time_based
结果随机读568K，随机写471k，延迟2毫秒左右
这个性能如果将服务端换成linux，会有进一步提升吗，或者还可以怎么优化smb性能，谢谢
详细结果：
fio-3.35
Starting 20 threads
Jobs: 20 (f=20): [r(20)][100.0%][r=2300MiB/s][r=589k IOPS][eta 00m:00s]
fiotest: (groupid=0, jobs=20): err= 0: pid=2100: Fri Jul 14 20:14:56 2023
read: IOPS=568k, BW=2219MiB/s (2327MB/s)(108GiB/50004msec)
slat (usec): min=3, max=18398, avg=12.28, stdev=19.87
clat (usec): min=78, max=24933, avg=1105.27, stdev=393.91
lat (usec): min=92, max=24956, avg=1117.55, stdev=395.27
clat percentiles (usec):
| 1.00th=[ 570], 5.00th=[ 709], 10.00th=[ 775], 20.00th=[ 857],
| 30.00th=[ 930], 40.00th=[ 996], 50.00th=[ 1057], 60.00th=[ 1123],
| 70.00th=[ 1205], 80.00th=[ 1287], 90.00th=[ 1434], 95.00th=[ 1598],
| 99.00th=[ 2212], 99.50th=[ 3326], 99.90th=[ 5538], 99.95th=[ 6456],
| 99.99th=[ 7963]
bw ( MiB/s): min= 707, max= 2336, per=100.00%, avg=2221.76, stdev=13.47, samples=1980
iops : min=181132, max=598114, avg=568764.03, stdev=3447.11, samples=1980
lat (usec) : 100=0.01%, 250=0.01%, 500=0.43%, 750=7.53%, 1000=33.22%
lat (msec) : 2=57.54%, 4=0.95%, 10=0.32%, 20=0.01%, 50=0.01%
cpu : usr=4.10%, sys=29.80%, ctx=0, majf=0, minf=0
IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=26.3%, 32=73.7%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=99.4%, 8=0.5%, 16=0.1%, 32=0.1%, 64=0.0%, >=64=0.0%
issued rwts: total=28403323,0,0,0 short=0,0,0,0 dropped=0,0,0,0
latency : target=0, window=0, percentile=100.00%, depth=32
Run status group 0 (all jobs):
READ: bw=2219MiB/s (2327MB/s), 2219MiB/s-2219MiB/s (2327MB/s-2327MB/s), io=108GiB (116GB), run=50004-50004msec
fio-3.35
Starting 20 threads
Jobs: 20 (f=0): [f(20)][100.0%][w=1855MiB/s][w=475k IOPS][eta 00m:00s]
fiotest: (groupid=0, jobs=20): err= 0: pid=6988: Fri Jul 14 20:19:32 2023
write: IOPS=471k, BW=1838MiB/s (1928MB/s)(89.8GiB/50003msec); 0 zone resets
slat (usec): min=3, max=5255, avg=15.80, stdev=19.99
clat (usec): min=2, max=12805, avg=1332.25, stdev=340.86
lat (usec): min=86, max=13002, avg=1348.06, stdev=342.88
clat percentiles (usec):
| 1.00th=[ 693], 5.00th=[ 906], 10.00th=[ 1004], 20.00th=[ 1123],
| 30.00th=[ 1205], 40.00th=[ 1254], 50.00th=[ 1319], 60.00th=[ 1369],
| 70.00th=[ 1418], 80.00th=[ 1500], 90.00th=[ 1631], 95.00th=[ 1778],
| 99.00th=[ 2409], 99.50th=[ 3064], 99.90th=[ 4686], 99.95th=[ 5211],
| 99.99th=[ 6587]
bw ( MiB/s): min= 756, max= 1934, per=100.00%, avg=1840.31, stdev= 9.04, samples=1980
iops : min=193578, max=495152, avg=471112.04, stdev=2313.67, samples=1980
lat (usec) : 4=0.01%, 50=0.01%, 100=0.01%, 250=0.01%, 500=0.14%
lat (usec) : 750=1.45%, 1000=8.06%
lat (msec) : 2=88.04%, 4=2.09%, 10=0.21%, 20=0.01%
cpu : usr=3.80%, sys=28.20%, ctx=0, majf=0, minf=0
IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=26.0%, 32=74.0%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=99.3%, 8=0.6%, 16=0.1%, 32=0.1%, 64=0.0%, >=64=0.0%
issued rwts: total=0,23530673,0,0 short=0,0,0,0 dropped=0,0,0,0
latency : target=0, window=0, percentile=100.00%, depth=32 |
发表于 2023-7-14 20:53:42