ReiserFS or ext3 on USB flash drive?

[Christopher Browne]

On 1/24/06, bassix-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org <bassix-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
> On 1/24/06, Christopher Browne <cbbrowne-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
> > On 1/24/06, bassix-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org <bassix-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
> > > Is this feasible? Does anyone know of any potential pitfalls from trying this?
> >
> > It may be feasible, but there is reason to disrecommend using a
> > journalled filesystem on flash, namely that journalling introduces a
> > "hot spot."
> >
> > Whether it's ReiserFS or ext3, you introduce a ~32MB section of the
> > device that is being *continually* updated every time any changes are
> > made on the filesystem.
> >
> > Flash devices can only cope with limited numbers of updates per
> > "sector", which means that you'll be diminishing the lifespan.
>
> That is a good point that I was interested in knowing. Will ext2 be
> OK? Will "running" an OS from flash memory greatly reduce it's
> lifespan due to this aforementioned limitation? Can someone illuminate
> how flash memory counts each "access"? Does access to every sector
> count as one access? Or, if I copy an 8MB file, does that count as one
> access? What if I copy 100 512K files (how many "accesses" would that
> be)?

I'm not certain what the granularity is, but the point is that every
*update* moves you towards the eventual failure of the device. 
*Access* doesn't do that; *updates* do it.

Copying an 8MB file would presumably involve:
 a) Writing 8MB of data on the filesystem;
 b) Writing a couple KB of data of filesystem metadata;
 c) If journalling, writing the couple KB of filesystem metadata in the journal

The failure of the device will most likely take place because:
 a) There's a section of flash being used for FS metadata that you have updated
      over and over 150K times, wearing out the flash memory, or
 b) You have burned through the journal 150K times, wearing out the
     flash memory used for that section of data.

Note that it probably *isn't* the 8MB that will "kill you"; that isn't
likely to be a "hot spot" on the flash device.  "Hot spots," updated
frequently, which are likely to either be metadata (e.g. -
file/directory entries in frequently updated directories) or journals,
are what will quickly reach "end of life."
--
http://www3.sympatico.ca/cbbrowne/linux.html
"The true  measure of a  man is how he treats  someone who can  do him
absolutely no good." -- Samuel Johnson, lexicographer (1709-1784)
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml