Btrfs With Linux 6.13 Delivers Performance Improvements & Other Features

[phoronix]

Phoronix: Btrfs With Linux 6.13 Delivers Performance Improvements & Other Features

Along with the early Bcachefs pull request for Linux 6.13, SUSE engineer David Sterba submitted all of the Btrfs file-system feature updates in an early pull request for this next kernel version. Btrfs is seeing new performance optimizations and other enhancements for Linux 6.13...

https://www.phoronix.com/news/Linux-6.13-Btrfs

[Kjell]

Would be really interesting to see how OpenZFS fares again BTRFS since it wasn't included in the last round (6.11)

Looks like the last ZFS benchmark was done in 2019? Huh

• https://www.phoronix.com/search/zfs
• https://www.phoronix.com/search/btrfs

[pkese]

In 5 years, when Bcachefs stabilizes and becomes reliable (and Btrfs gets RAID5)
will Btrfs be competitive (in terms of performance) against Bcachefs?

Bcachefs appears to have a very interesting architecture with its hybrid log-structured b-tree approach.


(I've been using Btrfs for over 10 years on over 10 machines, and I'm extremely happy with it.)

[varikonniemi]

In 5 years, when Bcachefs stabilizes and becomes reliable (and Btrfs gets RAID5)
will Btrfs be competitive (in terms of performance) against Bcachefs?

Bcachefs appears to have a very interesting architecture with its hybrid log-structured b-tree approach.


(I've been using Btrfs for over 10 years on over 10 machines, and I'm extremely happy with it.)

BTRFS will never be as fast as bcachefs, the question only is how big is the difference? Will it stay huge like today, or can btrfs shrink the gap to meaningless so all the questions shift to other aspects.

[cynic]

In 5 years, when Bcachefs stabilizes and becomes reliable (and Btrfs gets RAID5)
will Btrfs be competitive (in terms of performance) against Bcachefs?

Bcachefs appears to have a very interesting architecture with its hybrid log-structured b-tree approach.


(I've been using Btrfs for over 10 years on over 10 machines, and I'm extremely happy with it.)

I don't see any meaningful performance improvement coming for btrfs.
AFAICT limitations are due to its own architecture, so I think that minor implementation only improvments are to be expected.

so, I think that bcache will be faster.

BTW: I'm a happy btrfs user since ages as well.

[mdedetrich]

In 5 years, when Bcachefs stabilizes and becomes reliable (and Btrfs gets RAID5)
will Btrfs be competitive (in terms of performance) against Bcachefs?

Bcachefs appears to have a very interesting architecture with its hybrid log-structured b-tree approach.


(I've been using Btrfs for over 10 years on over 10 machines, and I'm extremely happy with it.)

I don't think its possible for btrs to get any meaningful performance improvements without changing the on disk format in an extremely fundamental way at which point you may as well make a new filesystem.

The RAID 5/6 issue write hole solution has already been worked on however in order to enable it you have to create a new filesystem (and copy the data over from the old one) because it requires a changed on disk format, you cannot upgrade a current bcachefs filesystem partition. This feature is also advertised as being fresh/new and requires testing.

[waxhead]

BTRFS will never be as fast as bcachefs, the question only is how big is the difference? Will it stay huge like today, or can btrfs shrink the gap to meaningless so all the questions shift to other aspects.

While you are probably right, you may be royally wrong too. The extent V2 tree can change that, but that being said I have not heard much about it in a while.

Personally I don't care if btrfs is a bit slower than any other filesystem as long as it is reliable. Of course enterprise businesses may have another opinion, but if speed is the issue they can throw more drives at the problem or faster hardware. Btrfs is about reliability and flexibility.

And while I am at it, I hope the next kernel release will feature the read policy changes being worked on. That seem to have the potential to really speed up things a lot (if you have many drives in your system). It does have its caveats of course but regardless it looks promising.

[gotar]

The extent V2 tree can change that, but that being said I have not heard much about it in a while.

AFAIK it was abandoned. There were several changes introduced and that's mostly it. Therefore it seems to be dead-end now.

[Dukenukemx]

I'm wondering why CachyOS defaults to Btrfs? It seems to be the slowest of all the file systems to choose from?

[deusexmachina]

Would be really interesting to see how OpenZFS fares again BTRFS since it wasn't included in the last round (6.11)

Looks like the last ZFS benchmark was done in 2019? Huh

• https://www.phoronix.com/search/zfs
• https://www.phoronix.com/search/btrfs

Definitely - with the OS running on it as well "zfs on root". And on SSDs, of course. At least one test with mirrored SSDs. Some test in high memory pressure would be great too.