View previous topic :: View next topic |
Author |
Message |
jancici Apprentice
Joined: 27 Jan 2004 Posts: 284 Location: Slovakia
|
Posted: Mon Sep 26, 2011 10:47 am Post subject: initialize Corsair Force series 3 SSD |
|
|
hi, I did buy this SSD : Corsair Force series 3 60GB
I did read few articles abut align partition and trying to setup my new drive ...
but I am not able to find what is :
- what is erase block size? is it 512KiB?
- what is block size? is it 4KiB?
I think that block size is 4kiB but I am not sure about erase block size.
thanks for your tips for size of erase block.
I am reading this article
http://www.nuclex.org/blog/personal/80-aligning-an-ssd-on-linux%22
and trying to follow Markus steps.
I want to crate boot partition with 32MiB size and root one with rest of size.
Well, let;s assume that 512KB is size of erase block.
here is output from fdisk with cylinder units
Code: | fdisk /dev/sdc
Command (m for help): u
Changing display/entry units to cylinders (DEPRECATED!)
Command (m for help): p
Disk /dev/sdc: 60.0 GB, 60022480896 bytes
32 heads, 32 sectors/track, 114483 cylinders
Units = cylinders of 1024 * 512 = 524288 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x1aadf6e2
Device Boot Start End Blocks Id System
/dev/sdc1 3 64 31744 83 Linux
/dev/sdc2 65 114483 58582528 83 Linux
|
and with sectors units Code: |
Command (m for help): u
Changing display/entry units to sectors
Command (m for help): p
Disk /dev/sdc: 60.0 GB, 60022480896 bytes
32 heads, 32 sectors/track, 114483 cylinders, total 117231408 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x1aadf6e2
Device Boot Start End Blocks Id System
/dev/sdc1 2048 65535 31744 83 Linux
/dev/sdc2 65536 117230591 58582528 83 Linux
Command (m for help): |
do you think it is correct?
from my point of view it is okay
start of partitions
2048 sectors * 512 B = 1048576 B
1048576 B / (512*1024) = 2 - this is okay
65536 sectors * 512 B = 33554432 B
33554432 B / (512*1024) = 32768 - this is also okay because it is integer
Last edited by jancici on Mon Sep 26, 2011 12:34 pm; edited 1 time in total |
|
Back to top |
|
|
jancici Apprentice
Joined: 27 Jan 2004 Posts: 284 Location: Slovakia
|
Posted: Mon Sep 26, 2011 12:33 pm Post subject: |
|
|
next step is creating FS
I decided for EXT4. I am not sure about parameters stride=32,stripe-width=32.
And my question is, how those parameters are connected with single HDD / SDD?
I am not using any kind of raid. What is chunk size in my case? and how to calculate stride and stripe-width?
thanks
I did run this for BOOT partition
Code: | mkfs.ext4 -b 4096 -O ^has_journal -L BOOT /dev/sdc1 |
and this for root
Code: | mkfs.ext4 -b 4096 -L ROOT /dev/sdc2 |
thank you for your comments, I would like to start to using that SSD soon, so I have some time to tune it and setup correctly.
I am planing to put /home and /var and /usr/portage to another HDD. I already using /tmp in RAM. |
|
Back to top |
|
|
HeissFuss Guru
Joined: 11 Jan 2005 Posts: 414
|
Posted: Tue Sep 27, 2011 4:29 am Post subject: |
|
|
You've got the main thing right. Fdisk set your block alignment to 1MB offset, so your partitions are aligned.
As far as EXT4 stride/stripe settings, I really doubt they'll make much difference. Assuming you have a 512K erase block size though, you'd want stride=1024, and then don't bother with stripe-width (if you had to set it to something, you'd set it to 1.)
You can test you partition with dd (assuming the main thing you're concerned with is write speed.)
Since you have a sandforce, you'll need to test with something random (zero would be way too fast.)
This creates a random 1GB file called /testfile.dd
Code: | dd if=/dev/urandom of=/testfile.dd bs=1M count=1024 conv=fsync |
Assuming you have 3+GB of RAM, the testfile.dd will be cached.
Rewrite it with:
Code: | dd if=/testfile.dd of=/realtestfile.dd bs=1M count=1024 conv=fsync |
You can re-run the second dd a couple of time to get an average.
For reads, you can just use 'hdaparm -tT /dev/sdc2' |
|
Back to top |
|
|
jancici Apprentice
Joined: 27 Jan 2004 Posts: 284 Location: Slovakia
|
Posted: Tue Sep 27, 2011 10:14 am Post subject: |
|
|
thanks
I did that test with 1GB file of random content
Code: | dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 12.6844 s, 84.7 MB/s
root@guliver /media/ROOT # dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 12.6039 s, 85.2 MB/s
root@guliver /media/ROOT # dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 12.6312 s, 85.0 MB/s |
85MB/s, well don;t know if this is good or not
so I did run same test on classic hdd
Code: | dd if=/realtestfile.dd of=/realtestfile_dva.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 9.82114 s, 109 MB/s
root@guliver /media/ROOT # dd if=/realtestfile.dd of=/realtestfile_dva.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 9.50574 s, 113 MB/s
root@guliver /media/ROOT # dd if=/realtestfile.dd of=/realtestfile_dva.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 10.0342 s, 107 MB/s |
this giving better numbers.
so, conclusion? should I tune something? thanks |
|
Back to top |
|
|
HeissFuss Guru
Joined: 11 Jan 2005 Posts: 414
|
Posted: Tue Sep 27, 2011 1:22 pm Post subject: |
|
|
Sequential write is about the only metric magnetic drives still come close to SSDs on.
I couldn't find stats on sequential incompressible write speed for your 60GB drive, but it looks like the 120 can do ~138MB/s. The 60 is probably a bit slower than that, depending on how they designed it.
http://www.anandtech.com/bench/Product/400
Are you using SATA 6, and if so, what type of controller? Are you using AHCI or legacy mode?
Can you repeat the test with the initial file created from /dev/zero instead of /dev/urandom? |
|
Back to top |
|
|
jancici Apprentice
Joined: 27 Jan 2004 Posts: 284 Location: Slovakia
|
Posted: Tue Sep 27, 2011 1:54 pm Post subject: |
|
|
thanks for answers and infos
creating ZERO file
Code: | dd if=/dev/zero of=testfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 10.1716 s, 106 MB/s |
test
Code: | root@guliver /media/ROOT # dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 4.48724 s, 239 MB/s
root@guliver /media/ROOT # dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 4.42384 s, 243 MB/s
root@guliver /media/ROOT # dd if=testfile.dd of=realtestfile.dd bs=1M count=1024 conv=fsync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 4.41844 s, 243 MB/s |
this looks much better sandforce is doing its job
I did run same test on classic hdd, it is slower in this case.
I have only SATA II on my mainboard, and using AHCI mode
Code: |
lspci
...
00:1f.2 SATA controller: Intel Corporation 82801JI (ICH10 Family) SATA AHCI Controller
... |
|
|
Back to top |
|
|
PM17E5 Apprentice
Joined: 10 Feb 2009 Posts: 241
|
Posted: Tue Mar 20, 2012 11:21 pm Post subject: |
|
|
Thank you, I found this thread to be very informative because I have a drive with the same exact specifications and I was implementing the same exact partition scheme and file system on my system. What are your preferred fstab options for root and boot? I'm assuming you have noatime and discard set in it? I was trying to make sense of the various alignment methods online but there are so many different ways to go about it, it looked like one could read for a good day or two and still find a differing opinion on how to do it right. |
|
Back to top |
|
|
jancici Apprentice
Joined: 27 Jan 2004 Posts: 284 Location: Slovakia
|
Posted: Wed Mar 21, 2012 8:32 am Post subject: |
|
|
meanwhile I did change ext4 to btrfs
here is fstab
Code: | /dev/sda1 /boot ext4 noatime 1 2
/dev/sda2 / btrfs noatime,nodiratime,discard 0 1 |
its good, that its helpfull ... |
|
Back to top |
|
|
PM17E5 Apprentice
Joined: 10 Feb 2009 Posts: 241
|
Posted: Wed Mar 21, 2012 1:02 pm Post subject: |
|
|
How is btrfs working out for you? I heard it doesn't have any recovery or repair software and that it's unstable? Is there a noticeable difference in performance? |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|