| 1 |
On Monday 03 August 2009 23:05:02 Paul Hartman wrote: |
| 2 |
> On Mon, Aug 3, 2009 at 3:22 PM, Grant<emailgrant@×××××.com> wrote: |
| 3 |
> > I know Linux systems aren't supposed to become fragmented, but I've |
| 4 |
> > also read that it can happen eventually. I'm on ext3. I've read that |
| 5 |
> > ext4 will have a defragmenter but that it doesn't have one yet. |
| 6 |
> |
| 7 |
> It's not that they aren't supposed to become fragmented, it is that |
| 8 |
> they try to avoid it. There is a big difference, and things like |
| 9 |
> streaming writes (downloads, bittorrents, etc) can cause extreme |
| 10 |
> fragmentation. |
| 11 |
> |
| 12 |
> The time-honored way of fixing this is "backup, delete, restore". In |
| 13 |
> my case my simple defragmenter is to move a file to tmpfs and then |
| 14 |
> move it back to the hard drive. I always do this to files I'm about to |
| 15 |
> burn to a CD/DVD to ensure the read speed is optimal. |
| 16 |
|
| 17 |
Until one day someone write a super-duper disk cache algorithm that delays |
| 18 |
writes safely, notices that you are putting back unmodified something you just |
| 19 |
deleted, then reverts "to be deleted" flag on the block pointers. meaning that |
| 20 |
nothing has changed. |
| 21 |
|
| 22 |
Lucky for us, I do not believe that such a driver has been written yet. |
| 23 |
Unlucky for us, I believe that such a driver is entirely possible. |
| 24 |
|
| 25 |
:-) |
| 26 |
|
| 27 |
-- |
| 28 |
alan dot mckinnon at gmail dot com |
Archives