Linux 4.16 Can Be A Lot Faster For Small I/O Activity

[onicsis]

Performance improvement for any FS? Or XFS only

[treba]

but only if Intel Meltdown and Spectre mitigation do not ruin your performance?

HDD will take advantages from this improvement?

We have
traditionally incremented that field on every inode data or metadata
change. Typically this increment needs to be logged on disk even when
nothing else has changed, which is rather expensive.

Sounds like this is about writing less to the disk. So HDDs should benefit a lot actually. And Meltdown is about cpu, not the drive. So This helps in cases where the drive is the limiting factor

Performance improvement for any FS? Or XFS only

This patch series converts existing accesses of the i_version field to a new API, and then converts all of the in-kernel filesystems to use it.

[Guest]

However, when somebody presents something like this, I'm very very suspicious of the "change" and hope the reviewers are suspicious as well.

That 244% figure is "best case" scenario, small writes that halve amount of work to do (writeback+inode update IIRC), with bigger writes you probably won't even notice the difference. Even though this is fairly common workload, I don't think that change is gonna be too noticeable, unless running on spinning drive perhaps.

[Azrael5]

Has this workaround been applied?