@scottalanmiller said:
4% is not a reasonable failure number. 3% is a best case for the best drives. External USB arrays don't get those drives. 3% is not achievable by those drives even under ideal (fixed, datacenter) conditions.
Who says 4% is not reasonable? Also, define failure. My understanding (please correct me if I'm wrong) is that a server will fail a drive if it reports a certain number of errors (or any errors?), but that doesn't mean the drive won't work or that the drive will have data loss does it? So many of my external drives may actually have errors (or have had temporary errors - the drive could fail to read, then try again and succeed) but they weren't errors that caused data loss (they weren't fatal errors) so were undetectable to me and didn't compromise my backups, even though a server would have failed them. To use our car analogy, a car can fail in lots of different ways without preventing you from completing your journey.
If 30% is realistic, then the probability of 17 drives (that's the number I get through a year) working without errors over the course of a year is 0.23%. Even with such a small sample size, that's a small enough number to make me doubt your figure. A 4% failure rate, gives me a 50% probability of being error free. Throw in dozens of laptops, and dozens of portable hard drives that I've looked after over the years, and 4% sounds a reasonable reflection of my experience, with a reasonably large sample size.
The figure that really matters is the probability of taking one of those drives after a site disaster and attempting a restore and having it fatally fail. I reckon that figure could be around 1%. I reckon human error is a much bigger risk to recovery failure than physical error (like the classic one of discovering you were using a cleaning tape as your backup).