[RC5] MacOS X - Hard disk sleep

Victor Denisov vdenisov at redline.ru
Tue Jun 12 15:25:42 EDT 2001


> I've actually heard that its worse to let your drive stop and start but
that may
> only be valid on older drives.  It had something to do with the head
bouncing every
> time it slows down or something.
>
> This is probably another urban legend though.

No, this is valid for modern drives, too (perhaps, even more valid than for
older models - see below).

I've worked for a small computer shop for a couple of years, doing repairs
of electronic components like monitors, FDDs, modems, etc. During this time,
I've observed that about 80% of HDD failures happen on HDD start-up when you
turn the power on.

I think the reason is twofold. First, it is the strain put on all components
(main engine, primarily) when they're trying to bring a drive on-line ASAP.
This is especially important for modern drives (which come on-line in about
5-10 seconds, older ones took up to 30 seconds to reach their nominal
rotation speed).

Another reason contributing to higher failure rate on drive power-on/off is
that drive heads actually ride on a very thin air cushion over the HDD
plate. This cushion is created by drive rotation. All kinds of special
measures are taken to ensure that heads will never touch the plate - this is
especially important when hard drive starts/stops. However, this measures do
fail from time to time, and you end up either with a broken head (bye-bye,
HDD), or with a nice long scratch on its surface (with data loss for the
appropriate track(s), but disk still (barely) useable). This is the main
failure reason for older drives, where such measures weren't as
reliable/sofisticated as in modern drives.

So, yes, it is _much_ better to never stop a hard drive - I've seen plenty
of servers working for a couple of years without a single power-down
thrashing their hard drives after a reboot, either planned or unplanned.
However,  sometimes pure efficiency has to give way to "political" reasons
(like, power conservation, reduced noise levels, etc).

With best regards,
Victor Denisov.

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list