> Would you trust a drive that fails on write ?
> What if it started to fail in the middle of writing a block ? Or it
> could
> write a journal entry but not the actual data, etc... ?

I wouldn't trust a drive that failed, period.  But - it doesn't work that
way.  It works like this:

In flash memory, each time you write to a block, you can detect how many
more writes are possible in that block.  So, it gives you the ability to
predict the failure, and prevent it before it happens.

Suppose (for the sake of easy numbers) each block is 512 bytes, and it is a
512M drive containing 1M blocks.  What they do is to actually build 1.2M
blocks inside the drive, but only tell the OS that there are 1M available.
The SSD drives secretly internally remap blocks when they drop below a
threshold, saving them *before* they fail.  The OS doesn't know anything
about it.

I had some talks with people in the SSD industry yesterday, who said, due to
consumer fears, they're overprovisioning and overcompensating.  Their goal
for now is to make SSD more reliable than spindle disks, because those fears
hurt their industry.  I must question whether there's any bias in that
feedback, and I must question how well they succeed in delivering that
stated goal, and I must question whether some usage pattern might defy their
expectations.  But I am somewhat reassured to know they're aware of such
fears, and making efforts to minimize it.


sdfdsfdsf

_______________________________________________
Tech mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to