Flash filesystems benchmarks

Flash filesystems benchmarks

[David Wagner]

	Hello,

As a contract with the CELF, we (Free Electrons) developed a flash
filesystem benchmarking suite ; it tests ubifs, jffs2, yaffs2 and
squashfs(over UBI+two different block transition layers). It has been
run on a Calao USB-9263 board and a IGEPv2 board and on 2.6.38, .39 and
3.0 kernels.

The results are shown as histograms and two different reportings exists:
 * filesystem comparison for a given board/kernel couple
 * kernel comparison for a given board/filesystem couple

All the details are to be found at:
	http://elinux.org/Flash_Filesystem_Benchmarks
Direct link to the test protocol:
	http://elinux.org/Flash_Filesystem_Benchmarks_Protocol
Link to the gitorious page:
	https://gitorious.org/ffs-benchmarks/ffs-benchmarks

Reviews of the test protocol, pointers to incoherent results or any
other comment is welcome.
We will keep posting more results.


	David Wagner

-- 
David Wagner, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

Re: Flash filesystems benchmarks

[Belisko Marek]

Hi,

On Thu, Jul 28, 2011 at 11:53 AM, David Wagner
<david.wagner@free-electrons.com> wrote:
>        Hello,
>
> As a contract with the CELF, we (Free Electrons) developed a flash
> filesystem benchmarking suite ; it tests ubifs, jffs2, yaffs2 and
> squashfs(over UBI+two different block transition layers). It has been
> run on a Calao USB-9263 board and a IGEPv2 board and on 2.6.38, .39 and
> 3.0 kernels.
>
> The results are shown as histograms and two different reportings exists:
>  * filesystem comparison for a given board/kernel couple
>  * kernel comparison for a given board/filesystem couple
>
> All the details are to be found at:
>        http://elinux.org/Flash_Filesystem_Benchmarks
> Direct link to the test protocol:
>        http://elinux.org/Flash_Filesystem_Benchmarks_Protocol
> Link to the gitorious page:
>        https://gitorious.org/ffs-benchmarks/ffs-benchmarks
>
> Reviews of the test protocol, pointers to incoherent results or any
> other comment is welcome.
> We will keep posting more results.
Great job ;). Maybe small README how to use scripts. I try to run
./performance-test.py but got en error because test.cfg doesn't exist.
Shouldn't there be input argument for specific board something like:
diff --git a/performance-tests.py b/performance-tests.py
index 78c83f0..33a4292 100755
--- a/performance-tests.py
+++ b/performance-tests.py
@@ -168,7 +168,12 @@ def main():
        global cur

        cfg = ConfigParser.ConfigParser()
-       cfg.read('test.cfg')
+
+       if (len(sys.argv) < 2):
+           print "board configuration parameter missing!!!"
+           return
+
+       cfg.read(sys.argv[1])

        board_name = cfg.defaults()['board_name']
        version = cfg.defaults()['version']


>
>
>        David Wagner
>
> --
> David Wagner, Free Electrons
> Kernel, drivers, real-time and embedded Linux
> development, consulting, training and support.
> http://free-electrons.com
> --
> To unsubscribe from this list: send the line "unsubscribe linux-embedded" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

thanks,

marek

-- 
as simple and primitive as possible
-------------------------------------------------
Marek Belisko - OPEN-NANDRA
Freelance Developer

Ruska Nova Ves 219 | Presov, 08005 Slovak Republic
Tel: +421 915 052 184
skype: marekwhite
icq: 290551086
web: http://open-nandra.com

Re: Flash filesystems benchmarks

[David Wagner]

	Hi,

On 28/07/2011 21:14, Belisko Marek wrote:
> Hi,
> Great job ;). Maybe small README how to use scripts. I try to run
> ./performance-test.py but got en error because test.cfg doesn't exist.
 > Shouldn't there be input argument for specific board something like:
 > [...]

Thanks for the feedback. You're right, READMEs are missing ; they'll come.

The current procedure, regarding the configuration file, is:
  * write a configuration file (like those already provided as examples)
  * cp it to test.cfg
  * launch performance-test.py
why the cp ? because it allows to tweak it without changing the original.
But it would certainly be best to use a proper argument parsing 
(argparse, like for gnuplotize_*.py and generate_graphs.py) and have an 
optional configuration file.


The material and documentation to reproduce the tests isn't fully 
available yet but will be very soon.  I can already give you a link to 
the debian root filesystem (to be exported through NFS) that was used ; 
however I can't guarantee yet that it will work. 
http://free-electrons.com/~david/pub/RFS-benchmarks.tar.lzma (1.1GB)


David.

--
David Wagner, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com