Fedora Server 22 Is Using The XFS File-System By Default

[deanjo]

Are you referring to the defaults? Btrfs for /home is a fully supported option. Seems like trusted to me. Of course they only enable only a supported subset of Btrfs.

They also allow ext2 (or many others). Because they allow it doesn't necessary mean they endorse it. That being said, I've lost more data to EXT3 / EXT4 over the years than I have ever with XFS or BTRFS (or even Reiser if you want to go way back).

[xeekei]

I would call this an urban myth. I recommend reading http://lwn.net/SubscriberLink/638546/c63a545bf06860b1/

Also, this change is only for Fedora Server.

If what Dave Chinner says is true; that 64-bit filesystems will run out of steam by 2025, then Btrfs better get stable soon. Otherwise we won't have much use of it for long.

[RahulSundaram]

They also allow ext2 (or many others). Because they allow it doesn't necessary mean they endorse it.

Their support documentation actually does endorse it by commercially supporting it.

[caligula]

If what Dave Chinner says is true; that 64-bit filesystems will run out of steam by 2025, then Btrfs better get stable soon. Otherwise we won't have much use of it for long.

Not exactly true. There are various segments for drives. If XFS becomes too outdated for large scale enterprise use, there are still small scale servers, home computers etc.

For instance if you know how hard drives work, the largest drives have multiple platters. The size growth per platter is not growing that fast. Here's one graph of the size trend in general

http://blogs.msdn.com/b/b8/archive/2011/11/29/enabling-large-disks-and-large-sectors-in-windows-8.aspx

The biggest drives are now 8 TB, it takes about 2 years to double the size at current rate (might slow down in the future). That means we hit 8 PB in 10 years and 8 EB in 20 years. There's plenty of time left for this kind of use. Now if you think about SSDs, add 5-8 years due to cost issues. Even then the faster spinning drives and more durable SSDs can use XFS. For example 15k rpm drives are at 600GB max now. That means extra 10 years to hit the limit. I'd expect XFS to become outdated in 2045.

[pal666]

Btrfs is a good pick already but its CoW nature makes it a bad fit for DB's

lolwut

v3.7 (Dec 2012)

per-file NOCOW

[gilboa]

We use XFS on servers that house billions of files (as in many-billions-of-files) as ext4 crocks at ~1-2B.
However, XFS is *very* slow when deleting files. I can delete ~50+K files / second on a high end server using ext4, with xfs managing only 10% of that (~4K files / second).
Its a shame ex4 never got upgrade to use 64/128bit I-nodes.

- Gilboa

[b15hop]

We use XFS on servers that house billions of files (as in many-billions-of-files) as ext4 crocks at ~1-2B.
However, XFS is *very* slow when deleting files. I can delete ~50+K files / second on a high end server using ext4, with xfs managing only 10% of that (~4K files / second).
Its a shame ex4 never got upgrade to use 64/128bit I-nodes.

- Gilboa

No one has mentioned ZFS. What's that like compared to XFS and Ext4?

[gilboa]

No one has mentioned ZFS. What's that like compared to XFS and Ext4?

Given the fact the ZFS is out-of-tree and will never be officially supported, using ZFS in production on Linux is plain crazy.

[b15hop]

Given the fact the ZFS is out-of-tree and will never be officially supported, using ZFS in production on Linux is plain crazy.

But doesn't ZFS have many other advantages regardless of out-of-tree?

[nanonyme]

But doesn't ZFS have many other advantages regardless of out-of-tree?

Afaik nothing Btrfs either doesn't have or isn't planned to have. ZFS is far more advanced than XFS or Ext4 though