File-System Benchmarks On The Intel X25-E SSD

[phoronix]

Phoronix: File-System Benchmarks On The Intel X25-E SSD

Late last month we looked at the Intel X25-E Extreme SSD on Linux. We ran this high-performance solid-state drive within a System76 Serval Notebook and compared its performance to a Seagate Momentus 7200.2 SATA HDD. During that testing we were just using the default EXT3 file-system, but now we have taken the Intel X25-E SSD for another spin as we looked at its performance when using the ReiserFS, JFS, XFS, EXT3, and EXT4 file-systems.

http://www.phoronix.com/vr.php?view=13600

[neuron]

I'd love to see reiser4 with cryptcompress on when your pulling in btrfs testing.
A basic mkfs.reiser4 -o create=ccreg40,compress=lzo1 for example.

It'll up the cpu usage, but it'd be interesting to see how much faster reiser4 is when seek times arn't involved.

[mirza]

Is it possible that files for compilation tests are all in cache already, so Kernel never touch file system code anyway, hence same results across the board? I don't see in TFA any measures taken for clearing cache between tests.

[_pma]

I wonder what could be the performance of the file systems specialized for solid state drives (JFFS2, YAFFS2, LogFS, UBIFS...). Any reason for not including them in this benchmark ?

[Gentooer]

Where's Reiser4?

[kraftman]

Where's Reiser4?

Far away from Linux kernel...

[_pma]

Far away from Linux kernel...

... deeply burried somewhere in california.

[tettamanti]

I wonder what could be the performance of the file systems specialized for solid state drives (JFFS2, YAFFS2, LogFS, UBIFS...). Any reason for not including them in this benchmark ?

Because they need direct access to the flash memory (via MTD), while SSD expose themselves as normal HDs (i.e. block devices).

[bulletxt]

Hey Michael, can you do an article about ATI Stream Software Development Kit (SDK)since it's the amd counterpart againts NVIDIA Cuda?

http://developer.amd.com/gpu/ATIStreamSDK/Pages/default.aspx

[kernelOfTruth]

... deeply burried somewhere in california.

if both is true - so what ?

it's still by far the best filesystem out there (btrfs doesn't count yet)

that criteria that a filesystem's included in the linux-kernel doesn't mean that it's superior compared to other filesystems

also the benchmarks are not fair, jfs doesn't support barriers from what I know

and ext3, ... have no support for extents / delayed allocation, ...

so when comparing those filesystems the differences have to be pointed out

also: WHERE ON EARTH IS SPACE EFFICIENCY / SPACE USAGE compared ?

especially with the early SSDs space is pretty precious and if you can store double compared with another filesystem it's worth gold ...

besides all of this ranting:


thanks very much for these benchmarks, they at least give an superficial possibility to compare the filesystems to each other