frequent automatic unloading on the heads ... failures of disks around 6-9 months after they are deployed ... generally have very high "load cycle counts" in the SMART data, and while inspecting one I found that there appeared to be tiny plastic particles ...
I find immediately I wonder:
Did the vendor report load cycle counts for drives tested til (accelerated) failure during design qualification?
Do those counts differ significantly from these counts seen correlated with death in the field?
"aggressive" power management
A newbie pitfall found in some t10.org standards is to code the power management to park at a fixed rate. Solve enough actual problems with such an app and eventually that fixed rate runs just above the fixed rate at which the app actually streams compressed data, for example. The natural implicit result is then a maximally accelerated test til failure: read, park, repeat, indefinitely.
Pat LaVarre
P.S. Yes, guesstimating disc/ drive life is a dark art. Yes, using a disc uses up the drive and the disc, by way of moving parts rather than decaying silicon. Yes, the physical damage implicit in the movement with friction of insert/ eject/ spin up/ spin down/ spin with heads on disc varies from one design to another.
