Right, so do usb flash drives hav ethe same wear leveling algorithms 
that the newer SSD's use?

aslo, I would want to put some things like /tmp into a ramdisk. also, 
could I disable the paging file? just thinking about reducing 
unnecessary writes.

Nathaniel C. Steele
Assistant Chief Engineer/Technical Director
WTRM-FM / TheCrossFM

On 10/11/2012 6:23 PM, Cowboy wrote:
> On Thursday 11 October 2012 05:37:26 pm Wayne Merricks wrote:
>> USB pen drives also generally have quite low write wear rates before the 
>> flash dies, its not a problem I've ever noticed over the years but I suppose 
>> if there is constant writes to the drive you might eventually burn out part 
>> of it.
>>
>> Having said that its probably not much less reliable than normal drives.
>>
>   Modern flash memory is typically good for 100K or so erase-write cycles.
>   Some of the newer flash are claiming 1M cycles, so yes, at some point they
>   will start reporting errors as the ROM wears out, but the real question
>   ( to which you allude ) is how many erase-write cycles will you be 
> subjecting
>   it to over its expected life ? ( erase is much harder on the ROM than write 
> )
>
>   In my experience as a data recovery guy, hard disks fail mostly due to 
> mechanical
>   reasons. Bearing failure in Western Digital, head crashes in Seagate, they 
> all
>   have their typical failure modes. These are functions of mechanical stress.
>   Flash doesn't suffer mechanical stresses, but the medium does "wear"
>   with number of cycles. The controller on solid state drives tends to use the
>   memory much like a conventional hard drive, where the data isn't erased.
>   The sectors are merely marked "available" and sit dormant until needed, so
>   even though a particular block ( and flash blocks are much larger. That's 
> how
>   they get speed over EEPROM ) has its wear limits, the controller by using 
> other
>   blocks can attempt to minimize wear on individual blocks, so a solid state
>   "hard drive" can easily last as long as spinning platters.
>   Longer, if the data on it is only a fraction of its capacity.
>

_______________________________________________
Rivendell-dev mailing list
[email protected]
http://lists.rivendellaudio.org/mailman/listinfo/rivendell-dev

Reply via email to