| 1 |
Neil Bothwick writes: |
| 2 |
|
| 3 |
> On Wed, 29 Feb 2012 00:25:00 +0100, Alex Schuster wrote: |
| 4 |
> |
| 5 |
> > > > Don't reduce it to 0, the lower this value is, the more |
| 6 |
> > > > fragmentation you will get. |
| 7 |
> > > |
| 8 |
> > > Why is that? I would have expected more usable space to reduce the |
| 9 |
> > > need for fragmentation. I routinely use 0 on non-system |
| 10 |
> > > filesystems. |
| 11 |
> > |
| 12 |
> > I read this often, and to me it seems to make sense. When a file |
| 13 |
> > system is nearly full, writing a last big file will make the file |
| 14 |
> > being cluttered along all those tiny places where some free space is |
| 15 |
> > still left. And this probably already happens to some extent before |
| 16 |
> > the filesystem is completely full. |
| 17 |
> |
| 18 |
> But if you set m > 0, the filesystem will become full sooner, so |
| 19 |
> fragmentation will begin sooner (for non-root processes). |
| 20 |
|
| 21 |
Uh, really? I wouldn't think so. With m > 0, there is much space left, in |
| 22 |
large contiguous chunks, even though the user cannot use it all. But there |
| 23 |
should be no difference between writing files in terms of fragmentation. |
| 24 |
The reserved stuff acts just like a quota, at least that's what I always |
| 25 |
thought. |
| 26 |
|
| 27 |
Wonko |
Archives