[Discuss] SSD

Bill Bogstad bogstad at pobox.com
Thu May 31 10:56:13 EDT 2012


On Thu, May 31, 2012 at 10:45 AM, Guy Gold <guy at the-golds.us> wrote:
> On Thu,May 31 10:31:AM, Edward Ned Harvey wrote:
>>
>> The MTBF of SSD's is sort of a black art.  When they first came out years
>> ago, they posted the same MTBF, but in actuality it was much worse because
>> windows kept writing the same disk block over and over, which is fatal to
>> SSD's.  But they fixed this problem with load leveling (or wear balancing)
>> in hardware in the SSD, mapping virtual blocks to physical blocks.
>
> I was told that the equivalent feature in Linux, that helps prolong the life of an
> SSD , started being available from Kernel 3.0 and above, true or false ?

I think what you are talking about and what Ed means are two different
things.   I think what Ed means is that the built-in controllers have
gotten better about moving data around so you don't overuse just one
erase block.    OS changes are probably the addition of support for
the "TRIM" command.   This command tells the drive that the OS no
longer cares what it gets from that data block in the future (i.e. the
files was deleted).   This can make the SSD controllers job easier
when trying to optimize wear leveling.   According to wikipedia:

http://en.wikipedia.org/wiki/TRIM#Operating_system_support

with the right kernel/filesystem this was available as early as 2.6.28
Linux kernel.   Precisely how well it does for a particular
kernel/filesystem mix I don't know.   You should follow up the
references to that wikipedia article if you really care.

Bill Bogstad



More information about the Discuss mailing list