Re: SSD Analyzer Tool -- MIDRANGE-L

OK Clearly spinny and SSD disks are different in many ways but there are some parallels. A spinny disk experiences soft errors - that is errors that can either be corrected or retried and completed successfully. An SSD can experience cell wear out and remapped. A Spinny disk can have sectors flagged bad and remapped. In each of these cases the drive wouldn't be flagged as failed or even failing in a single event. Given enough of these however and both drives would get flagged. This information certainly gets to the underlying code. Given enough they begin to show up in service tools and the message "Impending DASD failure" can be issued. Eventually even though the drive may not yet have failed it gets 'voted off the island' and marked FAILED and IBM service comes in and replaces it. This, of course, is where RAID 5, hot spare, RAID 6, MIRROR, (or a combination) save your data.

Remember too that SSD is a stepping stone for the next way to use solid state storage. It's convenient to package it as 'disk units' because we know how to interface to those and in this packaging it's easy to mount it, maintain it, and expand it. In the future it is likely to be connected much more directly to the system giving far more bandwidth and performance and of course better assessment of and direct knowledge of errors and other issues.

- Larry "DrFranken" Bolhuis

On 12/4/2011 10:31 AM, Joe Pluta wrote:

Is it your belief that a drive that had a cell fail rather than a chip would report it to the system? If not, then the distinction is moot - eventually enough cells will silently die that the drive will no longer be able to compensate. I agree that losing a chip is a catastrophic event. My point was only that the drive ought to let the system about degrading cells; it's pertinent information for systems administration.
As to having the extra cells, they are actually used for performance with added reliability being a side benefit. Writing a normal cell is a read
cell/update buffer/erase cell/write cell operation, but cells in the spare area are already erased so only a write is needed to store the data (the
cell being updated is mapped to the new cell& the old cell is put on the list of cells to garbage collect). So the extra cells aren't a reserved
area so much as cells that wind up scattered all over the memory chips over time. Depending on the drive's firmware& OS, the garbage collection of
used cells - where they're erased& prepped for future writing - occurs when the drive is idle or the OS issues a TRIM command.
This is a very cool concept. Clearly SSD technology has progressed significantly. Personally, I'm leaning towards a combination of SSD and
HDD on my next workstation. I'm just not sure yet that it's ready for the enterprise, especially at the current price point, but we'll see.

Joe