Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.os.linux.hardware > #3521 > unrolled thread

block oriented storage

Started byJohannes Koehler <jko@johannes-koehler.de>
First post2022-01-06 15:54 +0000
Last post2022-01-07 07:04 +0000
Articles 3 — 3 participants

Back to article view | Back to comp.os.linux.hardware


Contents

  block oriented storage Johannes Koehler <jko@johannes-koehler.de> - 2022-01-06 15:54 +0000
    Re: block oriented storage Marco Moock <mo01@posteo.de> - 2022-01-06 19:57 +0100
    Re: block oriented storage Henrik Carlqvist <Henrik.Carlqvist@deadspam.com> - 2022-01-07 07:04 +0000

#3521 — block oriented storage

FromJohannes Koehler <jko@johannes-koehler.de>
Date2022-01-06 15:54 +0000
Subjectblock oriented storage
Message-ID<slrnste452.srn.jko@IONNOKEFKO.johannes-koehler.de>
Much valued linux hardware hackers!

… its my first posting into your box. Because i am
lazy, i will shift a bit till later giving you more about 
myself @your togetherness, in the beloved txt form.

Short question...

I try to get more skills to know about the following relation:

(a) sectorsize@(hdd or ssd)
(b) blocksize@filesystem
(c) pagesize@(linux kernel or file)

Maybe, someone can give me a hint to a good manpage or
htmlpage.

thx and sincerely
- kefko

[toc] | [next] | [standalone]


#3522

FromMarco Moock <mo01@posteo.de>
Date2022-01-06 19:57 +0100
Message-ID<20220106195737.26c0dc1c@ryz>
In reply to#3521
Am Donnerstag, 06. Januar 2022, um 15:54:10 Uhr schrieb Johannes
Koehler:

> (a) sectorsize@(hdd or ssd)
Normally, one sector has 512 Bytes. There are now HDDs that have 4096
Byte sector, these are often called Advanced Format (AF).
https://en.wikipedia.org/wiki/Advanced_Format

[toc] | [prev] | [next] | [standalone]


#3523

FromHenrik Carlqvist <Henrik.Carlqvist@deadspam.com>
Date2022-01-07 07:04 +0000
Message-ID<sr8omr$pl6$1@dont-email.me>
In reply to#3521
On Thu, 06 Jan 2022 15:54:10 +0000, Johannes Koehler wrote:
> I try to get more skills to know about the following relation:
> 
> (b) blocksize@filesystem 

The block size depends on your filesystem. Example of a good man-page 
might be to study the -b flag at the man-page of mke2fs. To see the block 
size of an existing ext*fs you can try something like:

 dumpe2fs /dev/sda1

As you probably understand, it is prefered to have a block size which is 
a multiple of 1 or bigger of the sector size.

> (c) pagesize@(linux kernel or file)

The page size is usually 4 kB, you can check if that is really the case 
on your system by doing:

cat /proc/1/smaps | grep PageSize

Again, for swapping performance, it is prefered to have a page size which 
is a multiple of 1 or bigger of the sector size.

One simple way to test you raw disk performance with different block 
sizes is to do:

dd if=/dev/sda of=/dev/null bs=512

dd if=/dev/sda of=/dev/null bs=1024

dd if=/dev/sda of=/dev/null bs=2048

dd if=/dev/sda of=/dev/null bs=4096

dd if=/dev/sda of=/dev/null bs=8192

...

For RAID systems you might want to compare the performance with different 
stripe sizes.

regards Henrik

[toc] | [prev] | [standalone]


Back to top | Article view | comp.os.linux.hardware


csiph-web