Comment Re:reliability (Score 2) 183
I disagree with this faith in SMART to provide aqueduct warning. So does Google.
Out of all failed drives, over 56% of them have no count in any of the four strong SMART signals, namely scan errors, reallocation count, offline reallocation, and probational count.
We conclude that it is unlikely that SMART data alone can be effectively used to build models that predict failures of individual drives.
http://static.googleuserconten...
Google's analysis was of spinning hard disks, but I can not believe that SMART is somehow better at monitoring SSDs than spinning hard disks. I have personally had drives that pass every smart test and hard drive scan, but click and buzz in unnatural ways. Likewise, I have had SSDs suddenly fail that were, by all external tests before and after the failure, operating within expected parameters. It doesn't help that many SSDs have a habit of rendering the stored data inaccessible with no chance of recovery when they loose power. Spinning HD manufacturers solved that problem decades ago with self-parking read-write heads. Then again, there is no SMART test that's going to predict when an electrical component is going to suddenly burst into flames. (I've seen it happen!) With a spinning HD I could replace the logic board or send the disk out for recovery and get that data back, probably unscathed. With an SSD the odds would be in no-one's favor.
When it comes to SSDs, the PC vendors need to step up their game on data redundancy. SSD Raid 1 arrays or integrated backup to cheaper storage should be standard configurations.