Linux 4.0 Hard Drive Comparison With Six File-Systems

[Azrael5]

Im' curious about F2FS into an USB 3.0 or 3.1 memory flash device compared to the others.

[caligula]

I want to know the same thing. My guess is that this is an invalid test. Michael should look into it a bit more because it's not telling me enough. More details please.

Obvious question is, did he invalidate I/O caches?

[gregzeng]

NTFS tested is probably one of the Linux NTFS formats?

Phoronix: Linux 4.0 Hard Drive Comparison With Six File-Systems

It's been a while since last running any Linux file-system tests on a hard drive considering all of the test systems around here are using solid-state storage and only a few systems commissioned in the Linux benchmarking test farm are using hard drives, but with Linux 4.0 around the corner, here's a six-way file-system comparison on Linux 4.0 with a HDD using EXT4, Btrfs, XFS, and even NTFS, NILFS2, and ReiserFS.

http://www.phoronix.com/vr.php?view=21624

In the report and the comments following, I was very surprised that the extreme difference that NTFS has with the other file systems, in nearly every (all?) test.

On my multi-booting notebook computers, with Linux & Windows 8.1, all the DATA folders are on a one terabyte hard disk drive, with Win8.1 NTFS-compressed partitions. All the operating systems are on the motherboard's stock mSATA SSD. (Dell XPS-15, L521x). This NTFS compression is very different from that on Win7 & earlier, and different from the Lunux-created NTFS partitions. The latest version is less prone to data corruption.

The NTFS tested is probably one of the Linux NTFS formats? I've tried to locate the Phoronix benchmark for Windows, but cannot. Do you do this in Linux or Windows, via a virtual emulation?

[b15hop]

Can someone explain why ntfs random read test was so high? If it's due to caching, I thought most linux filesystems have cached pages.

I want to know the same thing. My guess is that this is an invalid test. Michael should look into it a bit more because it's not telling me enough. More details please.

[TheOrqwithVagrant]

Doesn't XFS suffer from 0-size corruption if power failures happen during writing?
ReiserFS does not. Other type of data corruption can occur but at least most of your data is still there.
With XFS you do need an UPS or make frequent backups.

That issue was fixed 10 years ago, in 2.6.17. I think it's about time we stop dragging that piece of ancient history into FS discussions.

[a7v-user]

So, if you needs something to replace reiserfs today, best bet is XFS. Down the road, it might be btrfs (and might not be too far down the road).

Doesn't XFS suffer from 0-size corruption if power failures happen during writing?
ReiserFS does not. Other type of data corruption can occur but at least most of your data is still there.
With XFS you do need an UPS or make frequent backups.

[caligula]

No, spinning rust does not optimize for fat32 to my knowledge. It does not need to. With flash there's an advantage to handling the superblocks specially, with a hd it's naturally faster if it's in the start of a partition.

FWIW, your spinning rust contains tens of megabytes of cache and command queues. It does not KNOW about fat32, but it can possibly do some optimizations that are beneficial to fat32. For example fat32 benefits from continuous blocks for files. Traditionally the fat table was loaded to ram so the bookkeeping task was quite fast.

[curaga]

And mechanical hard drives don't? Besides, it's still unbiased test, because neither the filesystem nor the computer itself are controlling the optimizations - only the drive's controller is doing the work here. That being said, if another filesystem is performing worse than FAT32 on such devices, that's still a problem.

No, spinning rust does not optimize for fat32 to my knowledge. It does not need to. With flash there's an advantage to handling the superblocks specially, with a hd it's naturally faster if it's in the start of a partition.

[schmidtbag]

To those posting about FAT32 being an unbiased comparison point: it is not. Most USB sticks, SD cards and SSDs have special optimizations for it in their firmware.

And mechanical hard drives don't? Besides, it's still unbiased test, because neither the filesystem nor the computer itself are controlling the optimizations - only the drive's controller is doing the work here. That being said, if another filesystem is performing worse than FAT32 on such devices, that's still a problem.

[budric]

Can someone explain why ntfs random read test was so high? If it's due to caching, I thought most linux filesystems have cached pages.