EXT4 With Linux 6.5 Will See Much Faster Parallel Direct I/O Overwrite Performance

[phoronix]

Phoronix: EXT4 With Linux 6.5 Will See Much Faster Parallel Direct I/O Overwrite Performance

Ted Ts'o has submitted all the EXT4 feature changes for the Linux 6.5 merge window. EXT4 this round is seeing various clean-ups, bug fixes, and other enhancements but there is one performance optimization worth calling attention to...

https://www.phoronix.com/news/Linux-6.5-EXT4

[MorrisS.]

From which version of the Kernel does the aforementioned regression affect Ext4?

[paulocoghi]

From which version of the Kernel does the aforementioned regression affect Ext4?

It could be that he is referring this regression by comparison, from EXT4 to XFS.

[deusexmachina]

I wish some ZFS devs cared about performance on NVMe (like on their own personal laptops) vs EXT4/F2FS/XFS... I'm sure it could be way closer, if we optimized for it and made a coherent profile by which testers could use to compare.

[fitzie]

It could be that he is referring this regression by comparison, from EXT4 to XFS.

I think it's likely ext4 vs ext4 regession, here's the original post the patch author made in February.

https://marc.info/?l=linux-ext4&m=167604118327461&w=2

of course, this is a corner case, as you aren't supposed to be sending unaligned IO if you care about performance (which would seem to be the case when you are doing direct io)

[curfew]

So it's fake news. There's no (new) performance increase, instead the patch is simply trying to regain some performance that was lost earlier when specs were changed.

[SleeepyKat]

I wish some ZFS devs cared about performance on NVMe (like on their own personal laptops) vs EXT4/F2FS/XFS... I'm sure it could be way closer, if we optimized for it and made a coherent profile by which testers could use to compare.

They care, but it's a slow process: https://www.youtube.com/watch?v=v8sl8gj9UnA

[mathieu.cossette]

ohhhhhhhh yes! August is coming, forget winter haha 🤣

[paulocoghi]

Thanks for the info!

I think it's likely ext4 vs ext4 regession, here's the original post the patch author made in February.

https://marc.info/?l=linux-ext4&m=167604118327461&w=2

[MorrisS.]

Thanks for the info!

So, the EXT4 regression has involved the 6.1 version of kernel and likely some previous kernels, it appears to be fixed since the 6.2 kernel.