EXT4 Scores A Nice Direct I/O Performance Improvement With Linux 6.3

[phoronix]

Phoronix: EXT4 Scores A Nice Direct I/O Performance Improvement With Linux 6.3

With the EXT4 file-system being quite mature at this stage, with many kernel cycles these days this widely-used file-system just sees bug fixes and other minor work. But for the newly-opened Linux 6.3 cycle, EXT4 is seeing a nice performance boost under certain conditions with direct I/O...

https://www.phoronix.com/news/Linux-6.3-EXT4

[M@GOid]

Well, since I love my data and hearing such news makes me nervous, how about leaving EXT4 alone and forking it to engage on such experimental gains with much more freedom? You can call it EXT5, release it to the schm... I mean, early adopters, and when the dust had settle down, us old farts can move to it in our own time.

[cynic]

Well, since I love my data and hearing such news makes me nervous, how about leaving EXT4 alone and forking it to engage on such experimental gains with much more freedom? You can call it EXT5, release it to the schm... I mean, early adopters, and when the dust had settle down, us old farts can move to it in our own time.

If you are concerned about stability you should probably consider running an LTS kernel instead of complain about innovation.

[avis]

Well, since I love my data and hearing such news makes me nervous, how about leaving EXT4 alone and forking it to engage on such experimental gains with much more freedom? You can call it EXT5, release it to the schm... I mean, early adopters, and when the dust had settle down, us old farts can move to it in our own time.

If you don't regularly backup, no FS, no matter how mature it is, is safe. Everything breaks. Except for the proton - that has to be yet observed.

[M@GOid]

If you are concerned about stability you should probably consider running an LTS kernel instead of complain about innovation.

I was not complaining, since I'm not against innovation. I was expressing my concerns, while offering a alternative. You can call it "constructive criticism" if it appease you. EXT4 is a critical piece of software for everybody, not only me.

Also, I do use a LTS kernel.

[cynic]

I was not complaining, since I'm not against innovation. I was expressing my concerns, while offering a alternative. You can call it "constructive criticism" if it appease you. EXT4 is a critical piece of software for everybody, not only me.

Also, I do use a LTS kernel.

the whole Linux kernel is a critical piece of software in pretty much every sector of our lives, not only ext4, still it has to advance.

the alternative you propose already exists: it's mainline kernel vs longterms.

Forking a filesystem for whatever minor or major improvement is totally unfeasible: we would have ext89 by now, and 89 branch to mantain.

[Jakobson]

I was not complaining, since I'm not against innovation. I was expressing my concerns, while offering a alternative. You can call it "constructive criticism" if it appease you. EXT4 is a critical piece of software for everybody, not only me.

Also, I do use a LTS kernel.

The patch itself is only 10 effective codeline changes in one single file. Very safe improvement and pretty easy to be reviewed.

https://lore.kernel.org/linux-ext4/[email protected]/T/#t

[timofonic]

Too much "experts" here...

[cynic]

Too much "experts" here...

is that before your post or after it?

[mathieu.cossette]

Like this news!