Originally Posted by parityboy @DuckieHo
So the cell damage would affect writes first? Also, is there an (at least vague) rule of thumb you can use to replace an SSD, if you can't get the needed info from it? Will the SMART spec be extended to include SSD features?
Cell damage pretty much only affects writes - once the cell is damaged you can no longer clear a cell to alter it's state. The cells remain readable throughout.
You don't really need to monitor the SSDs directly - your RAID card (or linux if under software RAID) can email you once the first cell fails to write, as it will trigger as a failed drive in the array. And since you're running RAID1 your system will continue operating as you replace the drive & rebuild - if you like at this point you can then replace the other drive as well (which is what I would recommend you do, as they will both be suffering very similar write patterns).
Originally Posted by tycoonbob
Where are you buying your SAS drives from? You can get 600GB 15K Seagate Cheetahs, as well as Hitachi Ultrastars for $200-250. HP and Dell branded drives aren't that much more either. Your comparison is also enterprise vs consumer. When using consumer SSDs with something that is getting over 800GB of writes a day, a SAS drive will last much longer. Enterprise SSDs are much more expensive, and in this case is all I would recommend. From the sounds of it, this isn't some little home hosted website. 800GB of Database writes per day, and over 200,000 users? Reliability and uptime is a must.
Where are you buying your SAS drives? In the UK I can't buy a warrantied HP 600GB SAS drive for close to that. And I'm buying them internally...
Edited by the_beast - 2/17/13 at 8:44am