On Wed, Aug 08, 2007 at 02:07:51PM +0200, Martin Steigerwald wrote:
>
> Hello!
>
> Today, I read about the impact of atime updates on ext3 and possible
> approaches on how to reduce that impact by using relatime, a modified version
> of relatime - Ingo Molnar posted a patch - or two different states of dirty
> inodes, atime dirty inodes which do not need to be flushed soon and dirty
> inodes with other changes which should be flushed soon[1].
Oh, that thread. Wake me when it's over, please?
> What would be the impact of atime versus noatime on XFS? What is the
> recommended setting for XFS? Are there any plans to implement relatime
> logic - maybe even the improved one by Ingo Molnar - into XFS?
relatime is a VFS layer construct - it should work with XFS right now.
> I just remounted my partitions on the notebook and my workstation with
> noatime
> for testing and I have the feeling that it indeed does help.
It might feel that way, but if you can't quantify it, it doesn't exist.
Joern Engel is the only person to present hard numbers in that thread,
and they showed pretty much no difference to performance....
Cheers,
Dave.
--
Dave Chinner
Principal Engineer
SGI Australian Software Group
|