Disk Capacity....

From: Paul Koning <pkoning_at_equallogic.com>
Date: Wed Jul 14 08:03:14 2004

>>>>> "Tom" == Tom Jennings <tomj_at_wps.com> writes:

 Tom> Areal comparisons don't hold, thermal expansion alone prevents
 Tom> just increasing capacity physically. It's not a coincidence the
 Tom> bigger drives are physically smaller.

I'm not sure why thermal expansion would matter; the servo can track
that trivially.

>From what I've heard, the reason for smaller drives is to allow them
to spin faster while maintaining acceptable size spindle motors, and
of course to reduce the seek distances. Access times for high end
disk drives are in the single milliseconds now -- that includes BOTH
an "average" seek and the time for 1/2 rotation.

Apparently, 15,000 rpm drives (high end fibre channel drives) have
lower capacity than slower drives not just because the areal density
is less, but also because the platter is smaller. You're getting what
is effectively a 2.5 inch drive packaged into a 3.5 inch chassis. The
reason the platter is that small is that you can't spin a 3.5 inch
sized platter at 15k rpm with acceptable motor power.

All this means that the modern way to large capacity is to combine
lots of (physically) small drives -- in other words, RAID. That also
gives you fault tolerance, and increased performance by putting more
heads to work.

 Tom> Encoding for modern data densities has long abandoned the
 Tom> simplistic schemes used in the deep past, too, and require tons
 Tom> of logic and software support to pull off.

Yes, and this is all packaged in just a small handful of hairy ASICs.
Modern data encode/decode schemes involve complex digital signal
processing at very uncivilized speeds...

           paul
Received on Wed Jul 14 2004 - 08:03:14 BST