On Mon, Sep 14, 2009 at 01:18:55PM -0500, Stephen Johnson wrote:

Flash again

(maybe  somebody  indeed  has found _HARD_ data about  wear  leveling,
something  like  "there are no more flash chips available  today  that
lack   an   in-hardware-implementation  of  wearleveling"   would   be
reassuring and down-right *dreamy*; if you'd success here, please drop
us a line:))

> When people think about the ramifications of writing over a TB of data
> to handheld PDA, they realize their fears were a bit silly. I'll admit I
> was worried about wearing out the flash of my IPAQ before I did the
> math. Even reflashing it a few hundred times, the IPAQ would be obsolete
> long before I wore out the flash in it.

IF  and ONLY IF there's some kind of wear-leveling active that spreads
out  hotspots  over  all the flash "pages" (or at least some  kind  of
reserve-page  substitution  or just even bad block marking).  Or  your
write   patterns  has  similar  characteristics.  Like  log-structured
file-systems  or  reflashing firmware. And for this case, the  TB  you
mention is easily more than enough.

Granted that the more expensive sata-ssd shoudl implement at least one
of these (proprietary mumble <unnamed> mumble) counter-measures.

But  hotspots  or swap hitting the same logical sector all  the  time?
Changing  512B all the time, requiring write/erase cycles all the time
on  the same cell. (what are the numbers here: 100K cycles, 1M at best
for the most expensive?)

Consider  USB sticks and directly soldered-in flash in small and still
expensive  gizmos. Gizmos naturally without any mention of anything at
all  in  available  docs wrt flash and wear-leveling? 

Consider  FAT and it's tendency of hot spots and it's unhappy marriage
to stupid bad old flash - I'd still suspect that as the reason for the
death my old creative muvo...

Is  it indeed safe nowadays to assume that some basic wear leveling is
implemented at HW-LEVEL in say HTC androids or Nokia's N900-to-be - so
I  can really place ordinary swap and logs on flash without having  to
consider/add work-arounds against hotspots?


Merely  finding hints online that wearleveling on flash has been known
in  the mid-nineties doesn't convince me this issue is indeed  ancient
history.  Doubly  so  for mass-market  consumer/prosumer-grade  stuff,
where builtin obsolescence is a marketing device.

still sceptical,
-- 
cu
Peter
[email protected]
_______________________________________________
Discuss mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to