Linux 3.14 Kernel File-System SSD Benchmarks

[Azrael5]

notably regression from 3.12 to 3.14 kernel. Could you test N2FS on both USB memory and SSD including new 3.15 kernel?

[grayhat]

So the F2FS drop in performance in 3.14 "seems to confirm a regression did take place and is still outstanding" - and particularly evident with HDDs as seen at http://www.phoronix.com/scan.php?pag...ltrabook&num=2

Now I was reading "Linux 3.14-rc5 Brings Stabilization To The New Kernel" and wondered if the situation has changed or not.

Can anyone shed some light on this issue? Michael, could you please post more recent disk performance benchmarks with kernel 3.14?

[guido12]

Any recent file system comparisons with USB drives and SD cards? Anyone have experience with using flashbench to optimize performance?

Thanks!

[Vim_User]

I don't get it why BTRFS does not ship with LZO compression enabled by default, it seems the performance gains are large and the downsides are a bit more CPU usage.

Maybe because there are still many SSDs with Sandforce controller that already does compression, without CPU overhead, so that LZO compression in BTRFS would not only be useless, but actually decreasing performance.

[ghexsel]

I get it for ARM devices, but for intel CPUs, I don't get it why BTRFS does not ship with LZO compression enabled by default, it seems the performance gains are large and the downsides are a bit more CPU usage. Other than SSD vs HDD, IO speeds and power consumption of these devices makes miniizing io something of a priority I would expect.

[shaurz]

F2FS is a great thing, but its biggest problem currently is fsck.f2fs. If you don't unmount a F2FS formatted flash card clearly, it soon gets corrupted. And fsck.f2fs is often incapable of repairing the damaged file system.

Ahh right. After my troubles with btrfs I will err on the side of caution.

[jpuhr]

F2FS

Is F2FS generally regarded as stable? I'm building a new SSD based system soon and I might consider it, with that write performance.

F2FS is a great thing, but its biggest problem currently is fsck.f2fs. If you don't unmount a F2FS formatted flash card clearly, it soon gets corrupted. And fsck.f2fs is often incapable of repairing the damaged file system.

[shaurz]

Is F2FS generally regarded as stable? I'm building a new SSD based system soon and I might consider it, with that write performance.

[jpuhr]

F2FS

Phoronix: Linux 3.14 Kernel File-System SSD Benchmarks
After last week delivering HDD file-system benchmarks on the in-development Linux 3.14 kernel, here are benchmarks of the Btrfs, EXT4, and F2FS file-systems from a solid-state drive.
http://www.phoronix.com/vr.php?view=19914

I use F2FS intensively, it's a great thing, but the biggest lack is still fsck.f2fs. When not unmounted clearly, the flash cards' file system gets corrupted soon; and fsck.f2fs is often not able to repair the damaged file system.

First, when running fsck.f2fs, assertions are triggered (assertions in release builds ?!?). If the assertions are deleted manually from the source code, fsck.f2fs succeeds, but the file system errors remain.

[phoronix]

Phoronix: Linux 3.14 Kernel File-System SSD Benchmarks

After last week delivering HDD file-system benchmarks on the in-development Linux 3.14 kernel, here are benchmarks of the Btrfs, EXT4, and F2FS file-systems from a solid-state drive.

http://www.phoronix.com/vr.php?view=19914