FreeBSD's embedded agenda

marty fouts mf.danger at gmail.com
Mon May 29 11:15:38 PDT 2006


On 5/28/06, James Mansion <james at wgold.demon.co.uk> wrote:
> >Experience suggests that wear leveling does matter in this market, but
> >that fairly simple wear leveling can be very effective.
>
> Can I ask why? I mmay be a  very bed person to talk about phones,
> because I have a 6-year-old Seimens I chose for battery life and
> the first thing I did was disable IrDa and WAP.  It doesn't have
> a camera. ;-)  And I've never sent a text.
>

Two reasons: First, NAND devices have a complicated wear behavior. The
more frequently you hit the same block, the lower the life expectancy
of the block.

Second, Because of the way NAND storage is used in smartphones, One
tends to accumulate fairly full devices, so the handful of data that
is frequently written tends to be limited to a small part of the
system. Many blocks see very little wear while a few blocks see a lot
of wear. This is exacerbated by the block size being large.

(smartphones tend to use NAND rather than NOR because of the cost difference)

Because of the relatively large block size, you have to do garbage
collection anyway, and it's not hard to do modest wear leveling while
you're doing that.


More information about the freebsd-small mailing list