Disks performance
abbas
8 Posts
March 14, 2022, 5:32 pmQuote from abbas on March 14, 2022, 5:32 pmFirst of all, thanks. I have a doubt, how could I know if the performance I am getting in the system with Petasan is good? I have seen the technical characteristics of the disks and together we think that it should have better IOPS and reading and writing speed.
I appreciate recommendations to check the performance I'm getting.
Thank you very much in advance.
First of all, thanks. I have a doubt, how could I know if the performance I am getting in the system with Petasan is good? I have seen the technical characteristics of the disks and together we think that it should have better IOPS and reading and writing speed.
I appreciate recommendations to check the performance I'm getting.
Thank you very much in advance.
admin
2,930 Posts
March 14, 2022, 6:30 pmQuote from admin on March 14, 2022, 6:30 pmThe performance varies with hardware. Have you already run benchmarks on the cluster from the benchmark pages ? can you run tests now or are you in production ? what is your current config ?
The performance varies with hardware. Have you already run benchmarks on the cluster from the benchmark pages ? can you run tests now or are you in production ? what is your current config ?
abbas
8 Posts
March 15, 2022, 6:17 pmQuote from abbas on March 15, 2022, 6:17 pmHi again,
Yes I can run the tests, I attach the results:
Test 4k IOPS Threads per Client 16 duration 1 min.
Results
Cluster IOPS
Write
Read
7596
83878
Test 4m throughput Threads per Client 16 duration 1 min.
Results
Cluster Throughput
Write
Read
2346 MB/s
2641 MB/s
Config: 3 nodes with 2 local storage service
POOL disk 8 NVME WD Red SN700 1000GB
type replicated - rbd - PG 512 - Size 2 - 4 NVME each node
If you need any additional information please ask me
Thanks again.
Best regards
Hi again,
Yes I can run the tests, I attach the results:
Test 4k IOPS Threads per Client 16 duration 1 min.
Results
Cluster IOPS
Write
Read
7596
83878
Test 4m throughput Threads per Client 16 duration 1 min.
Results
Cluster Throughput
Write
Read
2346 MB/s
2641 MB/s
Config: 3 nodes with 2 local storage service
POOL disk 8 NVME WD Red SN700 1000GB
type replicated - rbd - PG 512 - Size 2 - 4 NVME each node
If you need any additional information please ask me
Thanks again.
Best regards
admin
2,930 Posts
March 16, 2022, 11:52 amQuote from admin on March 16, 2022, 11:52 amRead iops is good but write is low. Throughput is good.
We recommend using enterprise class SSD, among other things they have high FUA/sync iops write speed. Not sure what the WD Red SN700 but it could have limited FUA/sync iops: You can test raw disk speed from PetaSAN blue console screen on a raw disk: it will destroy on data on it so do not use a running OSD or disk with data. The test will report the write sync iops. You could also use fio tool to do the same.
Back to the cluster benchmark test you did from ui: what was the % utlization of the disk and cpu ? It is also useful to run the test for 5 min and look at the utilization charts for cpu and disk utilization in dashboard.
Also if you have 4 disks per node you may get higher results by increasing the threads in the test from 16 to say 64
Read iops is good but write is low. Throughput is good.
We recommend using enterprise class SSD, among other things they have high FUA/sync iops write speed. Not sure what the WD Red SN700 but it could have limited FUA/sync iops: You can test raw disk speed from PetaSAN blue console screen on a raw disk: it will destroy on data on it so do not use a running OSD or disk with data. The test will report the write sync iops. You could also use fio tool to do the same.
Back to the cluster benchmark test you did from ui: what was the % utlization of the disk and cpu ? It is also useful to run the test for 5 min and look at the utilization charts for cpu and disk utilization in dashboard.
Also if you have 4 disks per node you may get higher results by increasing the threads in the test from 16 to say 64
Last edited on March 16, 2022, 11:52 am by admin · #4
abbas
8 Posts
March 16, 2022, 8:25 pmQuote from abbas on March 16, 2022, 8:25 pmTest 4k iops 64 threads 5 minutes:
Results
- Cluster IOPS
-
Write
Read
19764
125667
- Write Resource Load:
-
Node
Memory Util%
CPU Util%
Network Util%
Disks Util%
Actions
Avg
Max
Avg
Max
Journals
OSDs
Avg
Max
Avg
Max
NODO1
53
41
48
0
2
0
0
32
97
- Read Resource Load:
-
Node
Memory Util%
CPU Util%
Network Util%
Disks Util%
Actions
Avg
Max
Avg
Max
Journals
OSDs
Avg
Max
Avg
Max
NODO1
53
23
32
2
5
0
0
33
100
write Test fio in a nvme :
Starting 8 processes
Jobs: 8 (f=8): [w(8)][100.0%][r=0KiB/s,w=28.4MiB/s][r=0,w=7271 IOPS][eta 00m:00s]
xx: (groupid=0, jobs=8): err= 0: pid=186314: Wed Mar 16 21:24:07 2022
write: IOPS=10.9k, BW=42.8MiB/s (44.8MB/s)(430MiB/10046msec)
clat (usec): min=16, max=62065, avg=729.46, stdev=5984.30
lat (usec): min=16, max=62065, avg=729.53, stdev=5984.30
clat percentiles (usec):
| 1.00th=[ 23], 5.00th=[ 25], 10.00th=[ 26], 20.00th=[ 26],
| 30.00th=[ 26], 40.00th=[ 26], 50.00th=[ 26], 60.00th=[ 26],
| 70.00th=[ 26], 80.00th=[ 26], 90.00th=[ 28], 95.00th=[ 29],
| 99.00th=[52691], 99.50th=[56886], 99.90th=[60031], 99.95th=[61080],
| 99.99th=[62129]
bw ( KiB/s): min= 2480, max=40152, per=12.56%, avg=5500.46, stdev=8173.79, samples=160
iops : min= 620, max=10038, avg=1374.90, stdev=2043.50, samples=160
lat (usec) : 20=0.14%, 50=96.74%, 100=0.85%, 250=0.25%, 500=0.03%
lat (msec) : 2=0.04%, 4=0.06%, 10=0.54%, 20=0.21%, 50=0.06%
lat (msec) : 100=1.07%
cpu : usr=0.28%, sys=0.27%, ctx=109990, majf=0, minf=63
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 rwt: total=0,109968,0, short=0,0,0, dropped=0,0,0
latency : target=0, window=0, percentile=100.00%, depth=1
Run status group 0 (all jobs):
WRITE: bw=42.8MiB/s (44.8MB/s), 42.8MiB/s-42.8MiB/s (44.8MB/s-44.8MB/s), io=430MiB (450MB), run=10046-10046msec
Disk stats (read/write):
nvme3n1: ios=115/108688, merge=0/0, ticks=220/71520, in_queue=82204, util=100.00%
thanks
Test 4k iops 64 threads 5 minutes:
Results
- Cluster IOPS
-
Write
Read
19764
125667
- Write Resource Load:
-
Node
Memory Util%
CPU Util%
Network Util%
Disks Util%
Actions
Avg
Max
Avg
Max
Journals
OSDs
Avg
Max
Avg
Max
NODO1
53
41
48
0
2
0
0
32
97
- Read Resource Load:
-
Node
Memory Util%
CPU Util%
Network Util%
Disks Util%
Actions
Avg
Max
Avg
Max
Journals
OSDs
Avg
Max
Avg
Max
NODO1
53
23
32
2
5
0
0
33
100
write Test fio in a nvme :
Starting 8 processes
Jobs: 8 (f=8): [w(8)][100.0%][r=0KiB/s,w=28.4MiB/s][r=0,w=7271 IOPS][eta 00m:00s]
xx: (groupid=0, jobs=8): err= 0: pid=186314: Wed Mar 16 21:24:07 2022
write: IOPS=10.9k, BW=42.8MiB/s (44.8MB/s)(430MiB/10046msec)
clat (usec): min=16, max=62065, avg=729.46, stdev=5984.30
lat (usec): min=16, max=62065, avg=729.53, stdev=5984.30
clat percentiles (usec):
| 1.00th=[ 23], 5.00th=[ 25], 10.00th=[ 26], 20.00th=[ 26],
| 30.00th=[ 26], 40.00th=[ 26], 50.00th=[ 26], 60.00th=[ 26],
| 70.00th=[ 26], 80.00th=[ 26], 90.00th=[ 28], 95.00th=[ 29],
| 99.00th=[52691], 99.50th=[56886], 99.90th=[60031], 99.95th=[61080],
| 99.99th=[62129]
bw ( KiB/s): min= 2480, max=40152, per=12.56%, avg=5500.46, stdev=8173.79, samples=160
iops : min= 620, max=10038, avg=1374.90, stdev=2043.50, samples=160
lat (usec) : 20=0.14%, 50=96.74%, 100=0.85%, 250=0.25%, 500=0.03%
lat (msec) : 2=0.04%, 4=0.06%, 10=0.54%, 20=0.21%, 50=0.06%
lat (msec) : 100=1.07%
cpu : usr=0.28%, sys=0.27%, ctx=109990, majf=0, minf=63
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 rwt: total=0,109968,0, short=0,0,0, dropped=0,0,0
latency : target=0, window=0, percentile=100.00%, depth=1
Run status group 0 (all jobs):
WRITE: bw=42.8MiB/s (44.8MB/s), 42.8MiB/s-42.8MiB/s (44.8MB/s-44.8MB/s), io=430MiB (450MB), run=10046-10046msec
Disk stats (read/write):
nvme3n1: ios=115/108688, merge=0/0, ticks=220/71520, in_queue=82204, util=100.00%
thanks
Last edited on March 16, 2022, 8:29 pm by abbas · #5
abbas
8 Posts
March 16, 2022, 8:37 pmQuote from abbas on March 16, 2022, 8:37 pmcan I test raw disk speed from PetaSAN blue console screen on a raw disk? I can not find it. Thanks for everything
can I test raw disk speed from PetaSAN blue console screen on a raw disk? I can not find it. Thanks for everything
Last edited on March 16, 2022, 8:37 pm by abbas · #6
admin
2,930 Posts
March 17, 2022, 9:09 pmQuote from admin on March 17, 2022, 9:09 pmthe numbers are better now. do you see any saturation on the cpu % util or disk % util charts on dashboard ?
disk speed test is on the blue console menu, displayed on the node monitor not accessed from browser
the numbers are better now. do you see any saturation on the cpu % util or disk % util charts on dashboard ?
disk speed test is on the blue console menu, displayed on the node monitor not accessed from browser
Disks performance
abbas
8 Posts
Quote from abbas on March 14, 2022, 5:32 pmFirst of all, thanks. I have a doubt, how could I know if the performance I am getting in the system with Petasan is good? I have seen the technical characteristics of the disks and together we think that it should have better IOPS and reading and writing speed.
I appreciate recommendations to check the performance I'm getting.
Thank you very much in advance.
First of all, thanks. I have a doubt, how could I know if the performance I am getting in the system with Petasan is good? I have seen the technical characteristics of the disks and together we think that it should have better IOPS and reading and writing speed.
I appreciate recommendations to check the performance I'm getting.
Thank you very much in advance.
admin
2,930 Posts
Quote from admin on March 14, 2022, 6:30 pmThe performance varies with hardware. Have you already run benchmarks on the cluster from the benchmark pages ? can you run tests now or are you in production ? what is your current config ?
The performance varies with hardware. Have you already run benchmarks on the cluster from the benchmark pages ? can you run tests now or are you in production ? what is your current config ?
abbas
8 Posts
Quote from abbas on March 15, 2022, 6:17 pmHi again,
Yes I can run the tests, I attach the results:
Test 4k IOPS Threads per Client 16 duration 1 min.
Results
Cluster IOPS
Write Read 7596 83878 Test 4m throughput Threads per Client 16 duration 1 min.
Results
Cluster Throughput
Write Read 2346 MB/s 2641 MB/s Config: 3 nodes with 2 local storage service
POOL disk 8 NVME WD Red SN700 1000GB
type replicated - rbd - PG 512 - Size 2 - 4 NVME each node
If you need any additional information please ask me
Thanks again.
Best regards
Hi again,
Yes I can run the tests, I attach the results:
Test 4k IOPS Threads per Client 16 duration 1 min.
Results
Cluster IOPS
Write | Read |
---|---|
7596 | 83878 |
Test 4m throughput Threads per Client 16 duration 1 min.
Results
Cluster Throughput
Write | Read |
---|---|
2346 MB/s | 2641 MB/s |
Config: 3 nodes with 2 local storage service
POOL disk 8 NVME WD Red SN700 1000GB
type replicated - rbd - PG 512 - Size 2 - 4 NVME each node
If you need any additional information please ask me
Thanks again.
Best regards
admin
2,930 Posts
Quote from admin on March 16, 2022, 11:52 amRead iops is good but write is low. Throughput is good.
We recommend using enterprise class SSD, among other things they have high FUA/sync iops write speed. Not sure what the WD Red SN700 but it could have limited FUA/sync iops: You can test raw disk speed from PetaSAN blue console screen on a raw disk: it will destroy on data on it so do not use a running OSD or disk with data. The test will report the write sync iops. You could also use fio tool to do the same.
Back to the cluster benchmark test you did from ui: what was the % utlization of the disk and cpu ? It is also useful to run the test for 5 min and look at the utilization charts for cpu and disk utilization in dashboard.
Also if you have 4 disks per node you may get higher results by increasing the threads in the test from 16 to say 64
Read iops is good but write is low. Throughput is good.
We recommend using enterprise class SSD, among other things they have high FUA/sync iops write speed. Not sure what the WD Red SN700 but it could have limited FUA/sync iops: You can test raw disk speed from PetaSAN blue console screen on a raw disk: it will destroy on data on it so do not use a running OSD or disk with data. The test will report the write sync iops. You could also use fio tool to do the same.
Back to the cluster benchmark test you did from ui: what was the % utlization of the disk and cpu ? It is also useful to run the test for 5 min and look at the utilization charts for cpu and disk utilization in dashboard.
Also if you have 4 disks per node you may get higher results by increasing the threads in the test from 16 to say 64
abbas
8 Posts
Quote from abbas on March 16, 2022, 8:25 pmTest 4k iops 64 threads 5 minutes:
Results
- Cluster IOPS
Write Read 19764 125667
- Write Resource Load:
Node Memory Util% CPU Util% Network Util% Disks Util% Actions Avg Max Avg Max Journals OSDs Avg Max Avg Max NODO1 53 41 48 0 2 0 0 32 97
- Read Resource Load:
Node Memory Util% CPU Util% Network Util% Disks Util% Actions Avg Max Avg Max Journals OSDs Avg Max Avg Max NODO1 53 23 32 2 5 0 0 33 100
write Test fio in a nvme :
Starting 8 processes
Jobs: 8 (f=8): [w(8)][100.0%][r=0KiB/s,w=28.4MiB/s][r=0,w=7271 IOPS][eta 00m:00s]
xx: (groupid=0, jobs=8): err= 0: pid=186314: Wed Mar 16 21:24:07 2022
write: IOPS=10.9k, BW=42.8MiB/s (44.8MB/s)(430MiB/10046msec)
clat (usec): min=16, max=62065, avg=729.46, stdev=5984.30
lat (usec): min=16, max=62065, avg=729.53, stdev=5984.30
clat percentiles (usec):
| 1.00th=[ 23], 5.00th=[ 25], 10.00th=[ 26], 20.00th=[ 26],
| 30.00th=[ 26], 40.00th=[ 26], 50.00th=[ 26], 60.00th=[ 26],
| 70.00th=[ 26], 80.00th=[ 26], 90.00th=[ 28], 95.00th=[ 29],
| 99.00th=[52691], 99.50th=[56886], 99.90th=[60031], 99.95th=[61080],
| 99.99th=[62129]
bw ( KiB/s): min= 2480, max=40152, per=12.56%, avg=5500.46, stdev=8173.79, samples=160
iops : min= 620, max=10038, avg=1374.90, stdev=2043.50, samples=160
lat (usec) : 20=0.14%, 50=96.74%, 100=0.85%, 250=0.25%, 500=0.03%
lat (msec) : 2=0.04%, 4=0.06%, 10=0.54%, 20=0.21%, 50=0.06%
lat (msec) : 100=1.07%
cpu : usr=0.28%, sys=0.27%, ctx=109990, majf=0, minf=63
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 rwt: total=0,109968,0, short=0,0,0, dropped=0,0,0
latency : target=0, window=0, percentile=100.00%, depth=1Run status group 0 (all jobs):
WRITE: bw=42.8MiB/s (44.8MB/s), 42.8MiB/s-42.8MiB/s (44.8MB/s-44.8MB/s), io=430MiB (450MB), run=10046-10046msecDisk stats (read/write):
nvme3n1: ios=115/108688, merge=0/0, ticks=220/71520, in_queue=82204, util=100.00%thanks
Test 4k iops 64 threads 5 minutes:
Results
- Cluster IOPS
-
Write Read 19764 125667
- Write Resource Load:
-
Node Memory Util% CPU Util% Network Util% Disks Util% Actions Avg Max Avg Max Journals OSDs Avg Max Avg Max NODO1 53 41 48 0 2 0 0 32 97
- Read Resource Load:
-
Node Memory Util% CPU Util% Network Util% Disks Util% Actions Avg Max Avg Max Journals OSDs Avg Max Avg Max NODO1 53 23 32 2 5 0 0 33 100
write Test fio in a nvme :
Starting 8 processes
Jobs: 8 (f=8): [w(8)][100.0%][r=0KiB/s,w=28.4MiB/s][r=0,w=7271 IOPS][eta 00m:00s]
xx: (groupid=0, jobs=8): err= 0: pid=186314: Wed Mar 16 21:24:07 2022
write: IOPS=10.9k, BW=42.8MiB/s (44.8MB/s)(430MiB/10046msec)
clat (usec): min=16, max=62065, avg=729.46, stdev=5984.30
lat (usec): min=16, max=62065, avg=729.53, stdev=5984.30
clat percentiles (usec):
| 1.00th=[ 23], 5.00th=[ 25], 10.00th=[ 26], 20.00th=[ 26],
| 30.00th=[ 26], 40.00th=[ 26], 50.00th=[ 26], 60.00th=[ 26],
| 70.00th=[ 26], 80.00th=[ 26], 90.00th=[ 28], 95.00th=[ 29],
| 99.00th=[52691], 99.50th=[56886], 99.90th=[60031], 99.95th=[61080],
| 99.99th=[62129]
bw ( KiB/s): min= 2480, max=40152, per=12.56%, avg=5500.46, stdev=8173.79, samples=160
iops : min= 620, max=10038, avg=1374.90, stdev=2043.50, samples=160
lat (usec) : 20=0.14%, 50=96.74%, 100=0.85%, 250=0.25%, 500=0.03%
lat (msec) : 2=0.04%, 4=0.06%, 10=0.54%, 20=0.21%, 50=0.06%
lat (msec) : 100=1.07%
cpu : usr=0.28%, sys=0.27%, ctx=109990, majf=0, minf=63
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 rwt: total=0,109968,0, short=0,0,0, dropped=0,0,0
latency : target=0, window=0, percentile=100.00%, depth=1
Run status group 0 (all jobs):
WRITE: bw=42.8MiB/s (44.8MB/s), 42.8MiB/s-42.8MiB/s (44.8MB/s-44.8MB/s), io=430MiB (450MB), run=10046-10046msec
Disk stats (read/write):
nvme3n1: ios=115/108688, merge=0/0, ticks=220/71520, in_queue=82204, util=100.00%
thanks
abbas
8 Posts
Quote from abbas on March 16, 2022, 8:37 pmcan I test raw disk speed from PetaSAN blue console screen on a raw disk? I can not find it. Thanks for everything
can I test raw disk speed from PetaSAN blue console screen on a raw disk? I can not find it. Thanks for everything
admin
2,930 Posts
Quote from admin on March 17, 2022, 9:09 pmthe numbers are better now. do you see any saturation on the cpu % util or disk % util charts on dashboard ?
disk speed test is on the blue console menu, displayed on the node monitor not accessed from browser
the numbers are better now. do you see any saturation on the cpu % util or disk % util charts on dashboard ?
disk speed test is on the blue console menu, displayed on the node monitor not accessed from browser