Tangentially related, but still interesting: most people think hard drives have a bathtub curve when failure rate is plotted against time. Turns out this isn't true.

Schroeder and Gibson: "Contrary to common and proposed models, hard drive failure rates don’t enter steady state after the first year of operation. Instead failure rates seem to steadily increase over time."

http://www.pdl.cmu.edu/PDL-FTP/Failure/CMU-PDL-06-111.pdf