Stress Testing Hard Drives

David Spake dspake at gmail.com
Thu Jul 5 20:24:29 CDT 2007


On 7/2/07, Monty J. Harder <mjharder at gmail.com> wrote:
>> On 7/1/07, Jeffrey McCright <jmccright at hotmail.com> wrote:

> > Can anyone recommend a program or method of stress testing hard
>> drives?  I check memory/cpu with memtest86, but I would like some way to
>> stress test a hard drive.  Currently I dd urandom over it for a few days or
>> DBAN it, but I'm looking for something more thought out.  Preferably, a
>> program that can run on a live system so I'd just attach the drive to be
>> tested, and point the test at /dev/sdd or whatever dev it was on.
>> Preserving data on the drive is (obviously) not a concern.

On that suggested Ultimate Boot CD, the PowerMax tool for Maxtor
drives (2.09?) will perform a repetitive 'Stress Test' on internal
drives.  If I remember right, it will perform a complete write/read
test of every section of the drive for up to (I think) 30 times.
IIRC, it (unofficially) works with any drive, but you'll have to check
on it.

>
> Testing modern hard drives is complicated by the fact that the onboard
> controller manages defects internally.  When the drive writes a sector of
> data, it reads it back to verify that it can be done correctly.  If it
> can't, it locates a spare sector and tries to write there instead.  Once it
> finds a good spare sector, it records in its internal data structures that
> it has remapped the sector to the good location.

I saw some reasearch (January, Dec 06, ??) from Google and one of the
Univ. of California schools on hard drive analysis.  After testing
several hundred thousand drives from a variety of mfg's, they both
found that there was basically no way to tell what drive would fail
when, or in what fashion.  Google found that of the numerous (40?)
SMART parameters each HD companies tracks, with only 5 was there any
kind of statistical correspondance to failure.  By using those 5, they
were able to correctly predict upcoming failure in (get this) 40% of
their drives, thus leaving the other ~60% to some unknown cause.

> I've heard good things about SpinRite, but it's far from free as in beer,...

Spinrite is a fine program, which I'll vouch for (and paid for if for
nothing else to support the fine product he puts out).  GRC has been
producing this for many, many years and as far back as I can remember
it's always been head and shoulders above the competition.  I always
thought the program booted into soemthing other than DOS, but I could
be wrong there.

Dave


More information about the Kclug mailing list