The study, which used 10 different models of drives from Google's many data centres over "many millions of drive days", turned up one definitely unexpected nugget -- that the age of the drive, rather than the amount of usage it had seen, was the factor which correlated with high error rates.
