This chart shows the level of confidence I
would have on a particular hard drive being able to last x years, after
it has been in-shop for y days of testing and observation, if everything
is satisfactory and no new concerns are detected, after the problem is
deemed repairable and successfully rectified.
This chart is
based on a new drive less than 1-year old.
For each year of drive age, subtract 2.5%.
A minimal 1- to 3-week
observation period is recommended, for a meaningful
Basic assumptions: Drive is subject to 75% power-on
hours (for 7-day or less) and 50% power-on hours (for 10-day or
longer), with moderately heavy usage, mostly diagnostic exercises,
and occasionally, real world usage to get a few of its behavior
patterns. It is subject to reasonable daily thermal expansion-contraction
cycles, within a human hospitable environment.
According to manufacturers' declared Mean Time
Between Failure (MTBF) figures, a typical modern drive would expect
a 0.8% failure rate on an
annual basis. That is assuming artificial lab environment
under ideal conditions at all times. My empirical data suggests that
it is actually around 2%
in the field. That
ratio has been trending higher in the past 3 years, with the
advent of terabyte hard drives. With areal density on a steady rise
for 4 decades, we're now rapidly approaching the technological
barrier, with deteriorating reliability.
With world-wide hard drive manufacturing
consolidating, we now have only 5 makers. There is no discernable
difference across brands. There are however significant differences
across specific series, and sometimes even different capacity models
in the same series.
In the past decade, hard drives became the #1
hardware failure in computer systems. In the preceding 2 decades,
it was the power supply unit.