Jesse,
On Fri, 2 Jun 2006, Jesse Off wrote:
> > with flash because they do 3 device writes per software write
> operation...
> >
> > - write to journal
> > - do real data update
> > - rewrite journal record
> >
> > As you can see the journal area gets a lot of write hits too,
> hopefully
> > the wear levelling will help. But remember never use a journalling
> > filesystem with a floppy :-)
>
> Exactly. Though I believe EXT3 only journals metadata block writes so
> maybe its not too bad. If it truly was 3 writes for every real write
> we should be able to see that fairly obviously as a 3x slower file
> write speed though it may be a bit more complicated than that.
For an overview of of how jounralling filesystems work...
http://www.backupbook.com/03Freezes_and_Crashes/02Journaling.html
It's an interesting article with this quote....
"The price of using journaled systems
The problem with this is the problem of the journaled file system's
overhead. There are more I/O operations on the disk for each update and
most logging operations require synchronous writes. Also, since there is
much more writing and moving of blocks on the disk, fragmentation for the
system becomes a problem much faster than if journaling weren't installed
and running. The question becomes whether or not you want a faster or
safer system. This isn't a decision that should be made across the board,
but rather, on a case by case basis."
Of course that description assumes a normal diskdrive - you need to add
in the moves and changes the flash controller will make for various
purposes as well.
> A pet project I've had is to plug into the Linux block driver layer
> and actually do some accounting of number of block writes to some user
> accessible file in /proc. You could watch the flash write "load" by
> seeing how many writes accumulate during some sample period and
> therefore make predictions as to how long the flash should last at
> that write rate and optimize accordingly. You could even make
> objective comparisons between EXT2 and EXT3 to see if the extra
> stability of EXT3 outweighs the difference in flash lifetime on a
> project by project basis.
>
That is a brilliant idea. I've scratched my head off and on trying to
figure a way of getting that info. If you do do anything, here's one guy
would be interested.
cheers
Jim
------------------------ Yahoo! Groups Sponsor --------------------~-->
Everything you need is one click away. Make Yahoo! your home page now.
http://us.click.yahoo.com/AHchtC/4FxNAA/yQLSAA/CFFolB/TM
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/ts-7000/
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|