AVOIDING DATA DISASTER WITH
TODAY’S EVER-LARGER HARD DRIVES
There’s nothing like a hard
drive failure to bring
one’s backup strategy
into sharp focus, and
lately I seem to be surrounded by hard
drive failures. The first incident
occurred just before finals week, when I
lost my laptop’s year-old 500 GB internal
drive after a spontaneous crash. The
computer simply would not boot up
again, although I could hear it spinning.
A couple weeks later, a similar failure
struck the five-year old drive in a desktop
computer at home. Shortly after that I got
a call from a colleague inquiring whether
I could help him with an identical pair of
external drives that he’d owned for two
years, and which his computer no longer
recognized. It turned
out that each of the
hard drives involved
had suffered mechanical
failure, and no
amount of software
diagnostics, recovery
tools, or voodoo rituals could
bring them back to life.
Those of us who buy hard
disk storage for ourselves or
our organization are aware that
hard drives have simultaneously
increased in capacity and
decreased in price. It’s actually
becoming difficult to buy a hard
drive of less that 400 GB capacity. Most
are 750 GB and up, which means that losing
an entire drive can mean losing a lot
more content. Seagate, for example, has
announced their upcoming 3 TB drives
that could be available before the end of
the year. It also means that the 2 TB drives
will be significantly marked down. But
just how reliable can a huge drive be when
it retails for $70 including shipping?
IS QUALITY STILL ‘JOB ONE?’
Although drive capacity has increased and
manufacturing costs have dropped over
time, I wonder whether drive reliability
has actually decreased from what it was a
decade ago. A Carnegie Mellon University
report in 2007 debunked the concept that
MTBF (Mean Time Before Failure) was an
accurate predictor of hard drive longevity,
and indicated that drive failure may begin
much earlier than previously thought.
Moreover, reliance on on a drive’s selftesting
SMART function may not give
adequate warning, as Google pointed out
in its 2007 report. They found that drives
often needed replacement for
issues which SMART drive
status polling did not or could
not determine, and over half
of failed drives did not raise any significant SMART flags at all.
Three to five years is considered by professionals to be
the useful service life of a consumer-grade hard drive. Yet
it may be telling that several drive vendors have reduced
the length of their warranty period; Seagate’s used to be a
full five years on even their consumer drives, but it is now
three years for the newest models. Manufacturing missteps
which caused drives to fail prematurely have added
to customer concerns over decreasing service life. Simply
checking out customer ratings for hard drives on vendors’
websites reveals a decline in quality and an increase in
infant mortality—hard drives that die within 30 days of
use—along with more negative comments and ratings
from those who bought troubled drives.
AN ESSENTIAL INCONVENIENCE
If what I believe I am seeing is correct, then data backup is
more essential than ever given the
size of today’s hard drives.
I’m a fan of the 3-2-1 rule,
a strategy developed by
digital photographer
Peter Krogh in his
book “The DAM
Book: Digital Asset
Management for
Photographers”. This
rule states that one
should have three copies
of any file (that’s three different
devices, not three copies on
the same device), two different media types
(like hard drive and recordable DVD, for instance), and one
of the copies should be stored off-site. The off-campus
requirement can be met with an external drive that leaves
the office, or an Internet-based backup system such as
Carbonite or Mozy.
The ultimate answer to this entire issue may be to buy
large drives in pairs, then create a personal backup system
with them in a RAID 1 configuration. Meanwhile,
please excuse me, I need to go backup my backup.
info
CARBONITE:
carbonite.com
CARNEGIE MELLON
REPORT:
usenix.org/events/fast07/ tech/schroeder/schroeder_ html/index.html
GOOGLE’S HARD DRIVE
REPORT:
labs.google.com/papers/disk_ failures.pdfMozy mozy.com
PETER KROGH:
peterkrogh.com
Steve Cunningham is an assistant professor of practice
at USC’s Thorton School of Music.