Poor SSD performance

patrick7

Well-Known Member
Jan 25, 2018
32
1
48
30
Hi there

On my SSD servers, the SSD performance is really bad.
I test the performance with
Code:
dd if=/dev/zero of=test.img bs=4k count=262133 conv=fdatasync
(I'm aware that dd is not the best tool to test performance)


System 1 (SSD + ZFS & RAID1)
SuperMicro E300-8D
32 GB RAM
1x Samsung SSD 860 EVO 250GB (SATA)
1x Samsung SSD 970 EVO Plus 250GB (NVME)
ZFS Software RAID 1

Code:
(zrh1)root@vms1:~# dd if=/dev/zero of=test.img bs=4k count=262133 conv=fdatasync
262133+0 records in
262133+0 records out
1073696768 bytes (1.1 GB, 1.0 GiB) copied, 19.6898 s, 54.5 MB/s

Code:
(zrh1)root@vms1:~# pveperf
CPU BOGOMIPS:      35201.68
REGEX/SECOND:      215189
HD SIZE:           211.53 GB (rpool/ROOT/pve-1)
FSYNCS/SECOND:     2820.62
DNS EXT:           44.44 ms
DNS INT:           52.88 ms (mydomain.link)

Code:
(zrh1)root@vms1:~# fio --ioengine=libaio --direct=1 --sync=1 --rw=read --bs=4K --numjobs=1 --iodepth=1 --runtime=60 --time_based --name seq_read --filename=test.img

seq_read: (g=0): rw=read, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=1
fio-3.12
Starting 1 process
Jobs: 1 (f=1): [R(1)][100.0%][r=6014KiB/s][r=1503 IOPS][eta 00m:00s]
seq_read: (groupid=0, jobs=1): err= 0: pid=24093: Fri Dec 20 13:33:53 2019
  read: IOPS=1561, BW=6247KiB/s (6397kB/s)(366MiB/60001msec)
    slat (usec): min=488, max=51385, avg=625.70, stdev=211.97
    clat (usec): min=2, max=168, avg= 8.17, stdev= 9.65
     lat (usec): min=514, max=51423, avg=635.28, stdev=212.48
    clat percentiles (nsec):
     |  1.00th=[ 2992],  5.00th=[ 3024], 10.00th=[ 3056], 20.00th=[ 3088],
     | 30.00th=[ 3120], 40.00th=[ 3248], 50.00th=[ 3664], 60.00th=[ 3664],
     | 70.00th=[ 3856], 80.00th=[ 6048], 90.00th=[27264], 95.00th=[27520],
     | 99.00th=[28544], 99.50th=[29056], 99.90th=[42240], 99.95th=[69120],
     | 99.99th=[85504]
   bw (  KiB/s): min= 3080, max= 6752, per=99.98%, avg=6244.92, stdev=435.70, samples=119
   iops        : min=  770, max= 1688, avg=1561.20, stdev=108.91, samples=119
  lat (usec)   : 4=73.19%, 10=7.46%, 20=0.01%, 50=19.28%, 100=0.05%
  lat (usec)   : 250=0.01%
  cpu          : usr=2.39%, sys=97.31%, ctx=309, majf=8, minf=12
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=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=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=93702,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
   READ: bw=6247KiB/s (6397kB/s), 6247KiB/s-6247KiB/s (6397kB/s-6397kB/s), io=366MiB (384MB), run=60001-60001msec


System 2 (SSD + MDRAID 6 & LVM)
SuperMicro 5018D-FN8T
64 GB RAM
4x Samsung SSD 860 EVO 500GB (SATA)
1x Samsung SSD 970 EVO 500GB
Linux Software RAID 6 with Hot Spare

Code:
(zrh2)root@vms1:~# dd if=/dev/zero of=test.img bs=4k count=262133 conv=fdatasync
262133+0 records in
262133+0 records out
1073696768 bytes (1.1 GB, 1.0 GiB) copied, 11.2465 s, 95.5 MB/s

Code:
(zrh2)root@vms1:~# pveperf
CPU BOGOMIPS:      35202.88
REGEX/SECOND:      1939532
HD SIZE:           45.58 GB (/dev/mapper/vg0-host--root)
BUFFERED READS:    939.91 MB/sec
AVERAGE SEEK TIME: 0.14 ms
FSYNCS/SECOND:     143.51
DNS EXT:           26.50 ms
DNS INT:           31.67 ms (mydomain.link)

Code:
(zrh2)root@vms1:~# fio --ioengine=libaio --direct=1 --sync=1 --rw=read --bs=4K --numjobs=1 --iodepth=1 --runtime=60 --time_based --name seq_read --filename=test.img
seq_read: (g=0): rw=read, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=1
fio-3.12
Starting 1 process
Jobs: 1 (f=1): [R(1)][100.0%][r=93.6MiB/s][r=23.0k IOPS][eta 00m:00s]
seq_read: (groupid=0, jobs=1): err= 0: pid=1902: Fri Dec 20 13:31:53 2019
  read: IOPS=25.2k, BW=98.6MiB/s (103MB/s)(5913MiB/60001msec)
    slat (usec): min=5, max=145, avg= 8.02, stdev= 3.66
    clat (nsec): min=1182, max=73419k, avg=30545.47, stdev=122502.62
     lat (usec): min=25, max=73451, avg=38.70, stdev=122.86
    clat percentiles (usec):
     |  1.00th=[   21],  5.00th=[   21], 10.00th=[   21], 20.00th=[   22],
     | 30.00th=[   22], 40.00th=[   22], 50.00th=[   22], 60.00th=[   23],
     | 70.00th=[   23], 80.00th=[   24], 90.00th=[   27], 95.00th=[   98],
     | 99.00th=[  200], 99.50th=[  235], 99.90th=[  330], 99.95th=[  979],
     | 99.99th=[ 3294]
   bw (  KiB/s): min=54328, max=109736, per=100.00%, avg=100935.13, stdev=7821.72, samples=119
   iops        : min=13582, max=27434, avg=25233.78, stdev=1955.43, samples=119
  lat (usec)   : 2=0.01%, 4=0.01%, 10=0.02%, 20=0.15%, 50=93.68%
  lat (usec)   : 100=1.95%, 250=3.88%, 500=0.24%, 750=0.01%, 1000=0.01%
  lat (msec)   : 2=0.01%, 4=0.04%, 10=0.01%, 50=0.01%, 100=0.01%
  cpu          : usr=7.97%, sys=27.25%, ctx=1513659, majf=0, minf=13
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=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=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=1513794,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
   READ: bw=98.6MiB/s (103MB/s), 98.6MiB/s-98.6MiB/s (103MB/s-103MB/s), io=5913MiB (6201MB), run=60001-60001msec

Disk stats (read/write):
    dm-0: ios=1510636/964, merge=0/0, ticks=45956/17180, in_queue=63136, util=99.28%, aggrios=1513847/2791, aggrmerge=0/0, aggrticks=0/0, aggrin_queue=0, aggrutil=0.00%
    md0: ios=1513847/2791, merge=0/0, ticks=0/0, in_queue=0, util=0.00%, aggrios=303220/1937, aggrmerge=1085/7327, aggrticks=9787/2117, aggrin_queue=583, aggrutil=50.33%
  nvme0n1: ios=6/0, merge=0/0, ticks=85/0, in_queue=72, util=0.08%
  sdd: ios=378995/2546, merge=1448/8759, ticks=13169/3238, in_queue=688, util=50.33%
  sdc: ios=378757/2470, merge=1162/9324, ticks=11282/1132, in_queue=64, util=48.38%
  sdb: ios=379101/2424, merge=1345/9288, ticks=12530/3667, in_queue=1716, util=49.15%
  sda: ios=379242/2246, merge=1471/9268, ticks=11872/2548, in_queue=376, util=48.92%

System 3 (HDD + ZFS + RAID10)
SuperMicro 5017C-MTF
32 GB RAM
4x WDC WD1005FBYZ-01YCBB2
ZFS Software RAID 10

Code:
(fra1)root@vms2:~# dd if=/dev/zero of=test.img bs=4k count=262133 conv=fdatasync
262133+0 records in
262133+0 records out
1073696768 bytes (1.1 GB, 1.0 GiB) copied, 3.25517 s, 330 MB/s

Code:
(fra1)root@vms2:~# pveperf
CPU BOGOMIPS:      52800.16
REGEX/SECOND:      2101031
HD SIZE:           1760.22 GB (rpool/ROOT/pve-1)
FSYNCS/SECOND:     569.48
DNS EXT:           11.82 ms
DNS INT:           20.43 ms (mydomain.link)

Code:
(fra1)root@vms2:~# fio --ioengine=libaio --direct=1 --sync=1 --rw=read --bs=4K --numjobs=1 --iodepth=1 --runtime=60 --time_based --name seq_read --filename=test.img
seq_read: (g=0): rw=read, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=1
fio-3.12
Starting 1 process
Jobs: 1 (f=1): [R(1)][100.0%][r=159MiB/s][r=40.7k IOPS][eta 00m:00s]
seq_read: (groupid=0, jobs=1): err= 0: pid=2949: Fri Dec 20 13:42:00 2019
  read: IOPS=40.4k, BW=158MiB/s (166MB/s)(9472MiB/60001msec)
    slat (usec): min=20, max=291, avg=23.30, stdev= 3.37
    clat (nsec): min=602, max=33623, avg=731.08, stdev=378.63
     lat (usec): min=21, max=293, avg=24.16, stdev= 3.54
    clat percentiles (nsec):
     |  1.00th=[  644],  5.00th=[  652], 10.00th=[  660], 20.00th=[  668],
     | 30.00th=[  676], 40.00th=[  684], 50.00th=[  692], 60.00th=[  692],
     | 70.00th=[  700], 80.00th=[  708], 90.00th=[  756], 95.00th=[  948],
     | 99.00th=[ 1288], 99.50th=[ 1400], 99.90th=[ 8384], 99.95th=[ 8640],
     | 99.99th=[11328]
   bw (  KiB/s): min=151392, max=165504, per=99.98%, avg=161630.95, stdev=2146.14, samples=119
   iops        : min=37848, max=41376, avg=40407.71, stdev=536.53, samples=119
  lat (nsec)   : 750=89.81%, 1000=6.33%
  lat (usec)   : 2=3.61%, 4=0.06%, 10=0.19%, 20=0.01%, 50=0.01%
  cpu          : usr=7.10%, sys=92.91%, ctx=295, majf=8, minf=10
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=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=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=2424920,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
   READ: bw=158MiB/s (166MB/s), 158MiB/s-158MiB/s (166MB/s-166MB/s), io=9472MiB (9932MB), run=60001-60001msec

The (really old) server with only HDDs is much faster than the both SSD systems. I know that the RAID 1 / 6 performance is not as good as with RAID 10, but 54 / 95 MB/s is really poor for SSD RAIDs in my opinion.

All servers are almost IDLE. The servers were installed with ProxMox installer using the defaults.

Any ideas what the root cause could be?

Thanks and best regards
Patrick
 
Last edited:
You are right, dd if not a good benchmarking tool.
Have a look at our wiki where we have an article on how to benchmark disks: https://pve.proxmox.com/wiki/Benchmarking_Storage

Caution: don't give fio a disk as filename if you have data on it because it is destructive.
 
Update: It seems not to be disk related.
After some time, the system changes its CPU frequency to around 400 MHz. I have no idea why.
 
Yes, but it won't go over 500 Mhz after it was throttled down.
Any Ideas how I could find out the reason & how to turn it back?
 

About

The Proxmox community has been around for many years and offers help and support for Proxmox VE, Proxmox Backup Server, and Proxmox Mail Gateway.
We think our community is one of the best thanks to people like you!

Get your subscription!

The Proxmox team works very hard to make sure you are running the best software and getting stable updates and security enhancements, as well as quick enterprise support. Tens of thousands of happy customers have a Proxmox subscription. Get yours easily in our online shop.

Buy now!