OVH Community, your new community space.

Very bad perfomance on SSD Disk


phil21191
25/11/2015, 16h27
You're using the wrong parameters with DD to test sequential speeds.

Here is the results from my KS2

Code:
xxx@xxx:~# dd if=/dev/zero of=./test.tmp bs=4M count=1k
1024+0 records in
1024+0 records out
4294967296 bytes (4.3 GB) copied, 26.5686 s, 162 MB/s
xxx@xxx:~# dd if=/dev/zero of=./test.tmp bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 8.32105 s, 129 MB/s

And here are the same two tests ran on my home server which has a Kingston V300 SSD

Code:
xxx@xxx:[/tmp]: dd if=/dev/zero of=./test.tmp bs=4M count=1k
1024+0 records in
1024+0 records out
4294967296 bytes (4.3 GB) copied, 6.1623 s, 697 MB/s
xxx@xxx:[/tmp]: dd if=/dev/zero of=./test.tmp bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 3.25367 s, 330 MB/s
And here from the same homeserver on the ZFS pool

[code]

xxx@xxx:[/mnt/datastorage]: dd if=/dev/zero of=./test.tmp bs=4M count=1k
1024+0 records in
1024+0 records out
4294967296 bytes (4.3 GB) copied, 16.808 s, 256 MB/s
xxx@xxx:[/mnt/datastorage]: dd if=/dev/zero of=./test.tmp bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 15.0066 s, 71.6 MB/s

pch
12/11/2015, 09h36
Your result are perfectly in accordance with the SSD model used here.
From a smartctl result in the first page of this discussion they are Intel 320 40GB.

There is performance comparison of this SSD here: http://www.anandtech.com/show/4244/intel-ssd-320-review
The test itself is about a higher model but there is a table "Intel SSD Comparison":
For Intel SSD 320 40GB they give:
Sequential Write Performance : Up to 45MB/s
Sequential Read Performance: Up to 200MB/s

Patrick

Sebian
11/11/2015, 13h20
Citation Envoyé par myjticsolutions
But maybe all this servers with ssd disks are very slow.
It would seem like so. Some time has passed since you last asked, but I thought that I'd post my results, since nobody else did.

Here's my test procedure:

Code:
# Create a ramdisk with some random data for the tests
sudo mount -t tmpfs -o size=1200M none /mnt
dd if=/dev/urandom of=/mnt/testfile bs=1M count=1K conv=fdatasync
# This takes some time, I don't think it actually matters for the data to be random, so you can go with /dev/zero too.

# We've got our data ready under /mnt/testfile

# Drop any caches
echo 3 | sudo tee /proc/sys/vm/drop_caches

# Copy a file from ram to the ssd
dd if=/mnt/testfile of=/tmp/test bs=1M count=1K conv=fdatasync
# Output: copied 1073142708 bytes (1,1 GB), 22,5722 s, 47,5 MB/s

# Clear the caches again
echo 3 | sudo tee /proc/sys/vm/drop_caches      

# Copy a file from the ssd to ram
dd if=/tmp/test of=/mnt/testfile bs=1M count=1K conv=fdatasync
# Output: copied 1073142708 bytes (1,1 GB), 6,10429 s, 176 MB/s
My results are approximately:
KS-2 SSD: 45MiB/s write and 170MiB/s read
KS-3 (HDD): 145MiB/s write and 160MiB/s read

I'm guessing that the ssd's are ancient :s - take into account that they are 40GB - probably top of the line back in the day

I bet their random read/write is a lot better, compared to the hdds, but I don't know how to reliably test that

- Sebi

myjticsolutions
20/07/2015, 14h05
I bought this server one month ago, and I am still with a slow server. KS tech guys do not have solved the problem.

root@rescue:~# mkdir /ssd
root@rescue:~# mount /dev/sda1 /ssd
root@rescue:~# dd if=/dev/zero of=/ssd/test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 25.936 s, 41.4 MB/s



41.4Mb/s is very slow. In other KS server with no ssd disk I have a speed of three times faster.

When I got this server I was looking for a fast server, at least faster than the KS-1 and KS-2 servers. I guess you understand me:

KS-1 : 500 GB disk, IO 150 Mb/s
KS-2 SSD : 40 GB disk, IO 41 MB/s

If people know this info, nobody would want a server with a small disk and slow IO.

I would love to get other KS server with SSD disk and try it in order to know the IO speed, and so I would know if there is a problem with my current server, or there is a problem with KS-2 SSD servers. But I do not want to waste my money. I guess Kimsufi have tech guys who do that test, and if there are KS-2 SSD servers faster than my server is, they should give me one of them instead of my current slow server. But maybe all this servers with ssd disks are very slow.

nowwhat
09/07/2015, 17h48
Attention :
/dev/random blocs when there are no random number on stiock (I'm not kidding - these numbers take time to generate) and
/dev/urandom goes on with random numbers, but with less quality.

Getting '0' doesn't take time. Getting 'urandom' does take time.
Writing random '1' bits does take far more time on an SSD - and zero writng (if the space was already empty) take sless time.

All this explains a speed difference.
But not your 40,9 MB/s to 3,9 MB/s.

Strange that no-one posts here with results with their SSD-ON-KS-2 ....

I don't remember what my ZX81 did back then .... but I had one (also a C16, C64, an Atari 800XL )

myjticsolutions
09/07/2015, 17h06
Hi friend.
I think your results are OK. But please take a view of my results:

[root@ns ~]# dd if=/dev/zero of=/tests bs=64k count=16k conv=fdatasync
16384+0 registros leídos
16384+0 registros escritos
1073741824 bytes (1,1 GB) copiados, 26,2798 s, 40,9 MB/s
[root@ns ~]# dd if=/dev/urandom of=/tests bs=64k count=16k conv=fdatasync
16384+0 registros leídos
16384+0 registros escritos
1073741824 bytes (1,1 GB) copiados, 271,919 s, 3,9 MB/s
[root@ns ~]#

That is toooooo bad. I think my old spectrum had a better IO speed.

nowwhat
08/07/2015, 19h22
According to this: http://www.kossboss.com/linux---test...ives-raids-etc
I did a test :
dd if=/dev/urandom of=testfileR bs=8k count=10000; sync;
^C7628+0 records in
7627+0 records out
62480384 bytes (62 MB) copied, 4.01505 s, 15.6 MB/s

I guess random generating takes times - random bits writing also.

Your test:
dd if=/dev/zero of=/test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 8.11146 s, 132 MB/s

Writing 'zeros' goes fast !!

Note : I do not have a SSD, just plain 2 drives 'sda'/'sdb' - RAID 1.

myjticsolutions
08/07/2015, 17h38
Finally KS tech guys changed the hard disk. They took 12 days to do it.
But now the perfomance is even worse:

# dd if=/dev/zero of=/test bs=64k count=16k conv=fdatasync
16384+0 registros lei*dos
16384+0 registros escritos
1073741824 bytes (1,1 GB) copiados, 26,4056 s, 40,7 MB/s

So it is not a problem of the disk. What is happening?, I do not understand. I get an IO of around 120 MB/s in KS-1. I get better IOs in any VPS server.
I have installed Debian 7, Debian 6, Ubuntu 14, Centos 7, and always I get a low IO. Also with the rescue cd.

Could you please guys tell me what IO do you have in your SSD disks?

The IO is slow and the perfomance of the server is slow too. I have a web server with Drupal CMS.

myjticsolutions
02/07/2015, 12h55
Hello NOWWHAT.
Thank you for writing.
I remember two years ago that there was a problem with a disk and they changed it very soon and I did not have time to recover my data. I said "please I need my data" but of course they did not wait. Now I say once per day that I want they to change the disk or the motherboard, but they need to be safer and safer. Murphy's law I guess.
So, today I have written again asking to fix the problem with perfomance. I opened a new ticket as they have closed all my previous tickets. They close the tickets without answer.
Fabian, tell KS tech guys there is no data in the disk, they can change it, or destroy it.
Calm, I have to be patient, I guess 11 days waiting is not enough time to be angry. They need more time for a disk replacement.

nowwhat
01/07/2015, 21h20
Seems to me a double check before they blow away the drive.
They have to be SURE before they through away the drive and your data (if there is any data on it, that's NOT important).
You're up for a replacement ....

True : be patient .....

myjticsolutions
01/07/2015, 15h59
Still waiting for a solution.
In the last two days Kimsufi tech guys sent an email to me asking for backing up the data and asking for credential info. But nothing done yet.
That is boring.

myjticsolutions
30/06/2015, 12h27
Yesterday Kimsufi tech guys promised they would resolve my problem. I sent them again credentials to access my server. I sent again the results of the tests. It seems they like me to wait for days and days.

myjticsolutions
29/06/2015, 12h58
I confirm Kimsufi tech guys do not work on weekend.
After one week waiting for a solution, I am still waiting.
That is too bad. I have a hardware problem with my server. I paid for a SSD disk so the perfomance cannot be lower than it is for a normal HD. Kimsufi has not been able to change the disk or the motherboard after one week.
I am sad and disappointed with Kimsufi.

myjticsolutions
26/06/2015, 15h39
Hello Janus57.
A tech guy asked me about TRIM too. I did not know anything about it, but now I have googled a little.
As the server has a fresh OS installation and has been booted with Rescue CD, I think TRIM is not important now. I think TRIM would be important if I would be creating and deleting files.
Anyway I think the kernel has TRIM activated as I run the command " fstrim / " ( / is the root of the disk) from the console.
So, I actually think that is not the problem. I really think there is a hardware problem, either in hard disk or in motherboard. But tech guys are the only ones who can probe it.
Thank you for your interest.

janus57
26/06/2015, 14h38
Hello,

Tech guys work in weekend?
not on KS

Have you try to activate TRIM (don't know if rescue have it, so if rescue don't have it it the same as you server) ?

Support have no SLA on KS.

Cordially, janus57

myjticsolutions
26/06/2015, 14h26
Still waiting for a solution... 5 days without server. At least yesterday and today Kimsufi tech guys have sent emails to me.
I do not understand what the problem is.They said probably there was a problem with the kernel. But the system started with Rescue CD and it is very slow too.
I have others KS servers with non-SSD disks and they are faster than this one is.
Days fly, hours fly. And no server. I've got a headache now. Tech guys work in weekend?
If you, tech guys, don't find the problem with this server, why don't you give me other server with similar specs?, I guess it would works fine, I don't guess all SSD servers from Kimsufi have the same problem.
Fabian, help me man.

myjticsolutions
25/06/2015, 20h08
Cool !
Finally they have just answered my ticket. I have done other tests from console: "smartctl" and "hdparm" commands. I do not know what "fstrim" is, or if it is active on my kernel.
It is nice to have a support team to help customers. I have several kimsufi servers and I think it is the first time I need them.
I hope the problem is fixed very soon!

myjticsolutions
24/06/2015, 17h36
Hi.
NOWWHAT: Understood. And yes, that is a problem for me.
I sent tickets from client area, I sent forms from a contact page, and I write here. But no answer. I do not know what to do. It seems there are no tech-guys anywhere.
Thanks.

nowwhat
24/06/2015, 14h57
Citation Envoyé par myjticsolutions
.....
By the way, responding to NOWWHAT user. I do not understand these words: "Un SSD qui est moins rapide qu'un disque dur classique". SSD disks are faster than normal disks are. No doubt. Anyway, after I got the server, Kimsufi did a test and the IO speed was 248.03 MB/s. That is ok for a SSD disk. But in the tests I do, the speed is around 50MB/s. So I think Kimsufi does not do any tests, only "copy and paste" the report.
Sorry for that: I should have hit the "english" button first
What I said: "A SSD drive who's slower as a classic hard drive, that is a problem."

Did you post a ticket (that's NOT a froum post, there are NO KS tech-giuys on this forum) ?

myjticsolutions
24/06/2015, 13h57
Hi. Thank you for your answer. I have done what you said. The results of the hard disk test are:

---

smartctl 5.41 2011-06-09 r3365 [x86_64-linux-3.14.32-xxxx-std-ipv6-64-rescue] (local build)
Copyright (C) 2002-11 by Bruce Allen, http://smartmontools.sourceforge.net

=== START OF INFORMATION SECTION ===
Model Family: Intel 320 Series SSDs
Device Model: INTEL SSDSA2CT040G3
Serial Number: CVPR12230466040AGN
LU WWN Device Id: 5 001517 95958d543
Firmware Version: 4PC10302
User Capacity: 40,020,664,320 bytes [40.0 GB]
Sector Size: 512 bytes logical/physical
Device is: In smartctl database [for details use: -P show]
ATA Version is: 8
ATA Standard is: ATA-8-ACS revision 4
Local Time is: Wed Jun 24 12:40:52 2015 CEST
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

=== START OF READ SMART DATA SECTION ===
SMART overall-health self-assessment test result: PASSED

General SMART Values:
Offline data collection status: (0x00) Offline data collection activity
was never started.
Auto Offline Data Collection: Disabled.
Self-test execution status: ( 0) The previous self-test routine completed
without error or no self-test has ever
been run.
Total time to complete Offline
data collection: ( 1) seconds.
Offline data collection
capabilities: (0x71) SMART execute Offline immediate.
No Auto Offline data collection support.
Suspend Offline collection upon new
command.
No Offline surface scan supported.
Self-test supported.
Conveyance Self-test supported.
Selective Self-test supported.
SMART capabilities: (0x0003) Saves SMART data before entering
power-saving mode.
Supports SMART auto save timer.
Error logging capability: (0x01) Error logging supported.
General Purpose Logging supported.
Short self-test routine
recommended polling time: ( 1) minutes.
Extended self-test routine
recommended polling time: ( 1) minutes.
Conveyance self-test routine
recommended polling time: ( 1) minutes.
SCT capabilities: (0x003d) SCT Status supported.
SCT Error Recovery Control supported.
SCT Feature Control supported.
SCT Data Table supported.

SMART Attributes Data Structure revision number: 5
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE
3 Spin_Up_Time 0x0020 100 100 000 Old_age Offline - 0
4 Start_Stop_Count 0x0030 100 100 000 Old_age Offline - 0
5 Reallocated_Sector_Ct 0x0032 100 100 000 Old_age Always - 0
9 Power_On_Hours 0x0032 100 100 000 Old_age Always - 10802
12 Power_Cycle_Count 0x0032 100 100 000 Old_age Always - 41
170 Reserve_Block_Count 0x0033 100 100 010 Pre-fail Always - 0
171 Program_Fail_Count 0x0032 100 100 000 Old_age Always - 0
172 Erase_Fail_Count 0x0032 100 100 000 Old_age Always - 0
184 End-to-End_Error 0x0033 100 100 090 Pre-fail Always - 0
187 Reported_Uncorrect 0x0032 100 100 000 Old_age Always - 0
192 Unsafe_Shutdown_Count 0x0032 100 100 000 Old_age Always - 40
225 Host_Writes_32MiB 0x0032 100 100 000 Old_age Always - 278131
226 Workld_Media_Wear_Indic 0x0032 100 100 000 Old_age Always - 6907
227 Workld_Host_Reads_Perc 0x0032 100 100 000 Old_age Always - 16
228 Workload_Minutes 0x0032 100 100 000 Old_age Always - 648096
232 Available_Reservd_Space 0x0033 100 100 010 Pre-fail Always - 0
233 Media_Wearout_Indicator 0x0032 094 094 000 Old_age Always - 0
241 Host_Writes_32MiB 0x0032 100 100 000 Old_age Always - 278131
242 Host_Reads_32MiB 0x0032 100 100 000 Old_age Always - 56683

SMART Error Log Version: 1
No Errors Logged

SMART Self-test log structure revision number 1
Num Test_Description Status Remaining LifeTime(hours) LBA_of_first_error
# 1 Offline Completed without error 00% 10802 -
# 2 Reserved (0x80) Completed without error 10% 10752 -
# 3 Reserved (0x80) Completed without error 10% 10749 -
# 4 Reserved (0x80) Completed without error 10% 10749 -
# 5 Reserved (0x80) Completed without error 10% 10752 -
# 6 Offline Completed without error 10% 10749 -
# 7 Offline Completed without error 10% 10749 -
# 8 Short offline Completed without error 10% 10748 -
# 9 Reserved (0x20) Completed without error 10% 10748 -
#10 Reserved (0x20) Completed without error 10% 4765 -
#11 Reserved (0x20) Completed without error 10% 4752 -
#12 Reserved (0x20) Completed without error 10% 4752 -
#13 Vendor (0x78) Completed without error 10% 1098 -
#14 Vendor (0x78) Completed without error 10% 1085 -
#15 Reserved (0x20) Completed without error 10% 1085 -
#16 Reserved (0x20) Completed without error 10% 2 -
#17 Reserved (0x20) Completed without error 10% 0 -
#18 Reserved (0x30) Completed without error 10% 0 -

Note: selective self-test log revision number (0) not 1 implies that no selective self-test has ever been run
SMART Selective self-test log data structure revision number 0
Note: revision number not 1 implies that no selective self-test has ever been run
SPAN MIN_LBA MAX_LBA CURRENT_TEST_STATUS
1 0 0 Not_testing
2 0 0 Not_testing
3 0 0 Not_testing
4 0 0 Not_testing
5 0 0 Not_testing
Selective self-test flags (0x0):
After scanning selected spans, do NOT read-scan remainder of disk.
If Selective self-test is pending on power-up, resume after 0 minute delay.

---

I do not understand that. It seems everything is OK with the disk, but that is not true. The IO is very slow, slower than in a VPS with no ssd disk.
I think I will have to contact Paypal for them to give me back my money as Kimsufi does not resolve my problem. Bad luck. Very bad support. No support at all.

By the way, responding to NOWWHAT user. I do not understand these words: "Un SSD qui est moins rapide qu'un disque dur classique". SSD disks are faster than normal disks are. No doubt. Anyway, after I got the server, Kimsufi did a test and the IO speed was 248.03 MB/s. That is ok for a SSD disk. But in the tests I do, the speed is around 50MB/s. So I think Kimsufi does not do any tests, only "copy and paste" the report.

nowwhat
24/06/2015, 09h44
Une possibilité : bascule en rescue, fait les tests et envoi tout dans un ticket vers les services tech de KS.
Un SSD qui est moins rapide qu'un disque dur classique, il y a un soucis là.

myjticsolutions
23/06/2015, 14h31
Hi.
Thank you for your answer.
A fresh installation of Debian 8. Nothing more installed. The first command I put was "dd" because I felt the OS installation very slow.
I did the same test in Rescue Mode ant the results were the same.
I have been working with servers for several years, I am a proffesional, so I know there is something bad with the SSD disk or with the mainboard. But I need somebody from Kimsufi test it. Till know nowbody answered my tickets.
Any advices?
Thank you.

nowwhat
23/06/2015, 14h11
Hi,

What OS ?
Other tasks were running ?
You tested in rescue mode ? (if not, why not ???????)
Etc etc.

myjticsolutions
23/06/2015, 13h23
Hi.

On Sunday I got a server "KS-2 SSD". It has a SSD disk and I thought it would have a high IO speed. But please watch my results:

# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 registros leídos
16384+0 registros escritos
1073741824 bytes (1,1 GB) copiados, 22,4051 s, 47,9 MB/s

When I bought the server, I got a hardware test saying that the IO speed was 248.03 MB/s, very far from my results.

I have opened several tickets in these days, but no answer. What can I do?, should I open a dispute in Paypal?, I would love the problem is fixed as I need the SSD disk perfomance. I am losing money.

I hope somebody can help me.
Thank you.