[Gllug] Two Queries

Russell Howe rhowe at wiss.co.uk
Sun Jul 4 01:05:08 UTC 2004


On Sat, Jul 03, 2004 at 11:25:26PM +0100, Christopher Currie wrote:
> Neither O/T, I hope, but not related to recent discussions:

> 2. Is there any evidence that ReiserFS imposes more physical wear and tear 
> on hard drives than other filesystems?

With a journaling fs, the journal tends to get accessed a lot, and there
has been some speculation that this may cause excessive wear to the part
of the disk with the journal stored on it, but when this was brought up
on linux-xfs, nobody came forward with anything but theories, not even
anecdotal evidence, much less statistics.

Since there is (or should be) no contact between the head and the drive
surface, I can't imagine that with a regular hard disk it would matter
much how often you read/wrote to a particular point, unless the
substrate started to lose its ability to retain magnetism, which I would
assume is unlikely...

With floppy drives, the disk surface is pinched between the heads
(double sided disk), and so accessing the same portion of the disk
causes wear (as with tape systems).

I'd expect HD wear to be more akin to CD wear, likely to suffer
mechanical or chemical failure before parts of the disk wear out...

Now, the only thing which would occur to me is that writing to both the
journal and the main filesystem could cause the head to seek more (but
then it would with something like FAT or any other filesystem which
holds metadata in a single place - I believe ext2 and most other *nix
fs'es would be similar)

JFFS (and I think there's JFFS2) was mentioned - flash memories stop
working after a certain number of rewrites, so I assume JFFS spreads
modifications as evenly as it can without destroying performance...

-- 
Russell Howe       | Why be just another cog in the machine,
rhowe at siksai.co.uk | when you can be the spanner in the works?
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list