2002-10-08 19:35:56

by Andrew Morton

[permalink] [raw]
Subject: Re: The reason to call it 3.0 is the desktop (was Re: [OT] 2.6 not 3.0 - (NUMA))

Theodore Ts'o wrote:
>
> On Mon, Oct 07, 2002 at 07:59:26PM -0700, Andrew Morton wrote:
> > In the testing which I did, based on Keith Smith's traces, the
> > current code really isn't very effective.
> >
> > What I did was to run his aging workload an increasing number of
> > times. Then measured the fragmentation of the files which it
> > left behind. I measured the fragmentation simply by timing
> > how long it took to read all the files, and compared that to
> > how long it took to read the same files when they had been laid
> > down on a fresh fs.
>
> What access pattern did you use when you read the files? Did you
> sweep through filesystem directory by directory, or did you use some
> other pattern (perhaps random)?

Well this is all rather dim in my memory, so the confidence level
is drooping. I am sure that the timing was a single find | xargs cat
thing. I also know that I investigated whether the increased time
was due to the metadata access or the data access. I _think_ it was
mainly metadata.

But it all needs to be redone, really.

> ...
> > Maybe a mount option? But I think the current algorithm should
> > default to "off".
>
> How about a mount option with the possible values: "fast", "slow",
> "hinted", and "auto", with the default being "auto" or "hinted"?
> (Where hinted utilizes user-space hints, and "auto" utilizes
> user-space hints if present, plus some of the so-called ugly
> hueristics which you had discussed?)
>

Well the current Orlov patch will spread top-level directories,
so as long as /home is a mountpoint, we're fine.

For more generalality, yes, I think a new chattr flag on the
parent directory which says "spread my subdirectories out"
would be a good solution.


2002-10-08 22:01:00

by Rob Landley

[permalink] [raw]
Subject: Re: The reason to call it 3.0 is the desktop (was Re: [OT] 2.6 not 3.0 - (NUMA))

On Tuesday 08 October 2002 03:39 pm, Andrew Morton wrote:

> Well the current Orlov patch will spread top-level directories,
> so as long as /home is a mountpoint, we're fine.
>
> For more generalality, yes, I think a new chattr flag on the
> parent directory which says "spread my subdirectories out"
> would be a good solution.

Individual sysadmins may not use it much, but getting distributions to put it
in their install/upgrade software isn't too unlikely...

Rob