> >> I wonder if it might not be more useful to only spin up drives
> >> when they are actually needed, be it at mount time or even only
> >> upon actual access. This might be especially useful for drives
> >> used mainly as backup storage (e.g. to keep source trees that
> >> only need to be accessed every other week)...
> >
> > I guess I am not really in favor of spinning down disks all the
> > time. Most of the wear and tear that decreases the lifetime of a
> > disk is at startup and shutdown, so unless there are power management
> > reasons (i.e. on a laptop), I think it is probably better to simply
> > leave the disks spinning all the time.
> >
> > In principle, it would probably not be that harmful to perform
> > spinup/spindown on mount/unmount. Having an idle timer seems like a
> > dangerous idea to me, as you could wear out a disk in a real hurry.
>
> Well, what about a _long_ timeout?
>
> Always spindown after 26 hours without an access.
> Spindown after 1 hour if the last access was umount.
>
> (the 26 is in case you do something one a day and are late)
It seems to me that the question should be whether there is a good reason
to spin the drive down IN THE PARTICULAR CIRCUMSTANCES. The
crash-rip-grind-tear-scrape of the spindle bearings during a start-stop
will almost certainly subtract from the remaining life of a drive. The
heads will also give up a few of their N lives. But with a properly cooled
drive, there is very little wear and tear during normal operation.
In the special cases of laptops, where people will trade their first-born
children for a few more minutes of operation, then it definitely is an
economic trade of giving up drive-life instead. Similarly the case given
here where the drive was noisy enough to cause potential medical and sanity
problems.
But for the rest of us, I think that the way to a long-lived machine is to
let it run 24x7, especially the drives. Monitors are a different story as
there is definitely a life mechanism in the CRT phosphor. Of course, there
is also a tradeoff here with the inrush current into the cold filament of
the CRT being capable of literally tying physical knots in the cathode
structure. (I spent most of my life designing CRT display systems for
military equipment.)
Harvey