Back in the days when you had to
low level format
your
hard disk every now and then, and reading
floppies was
a time critical task that was almost too much for the cpu to handle,
and hardware track
buffers weren't an integral part of your disk drive,
it was interesting to set the sector interleave on your disks.
The interleave factor set the physical geometry order of the sectors
on the disk. (Example: an 8 sector track)
An interleave factor of '1' would put physically
neighboring sectors in consecutive order (0 1 2 3 4 5 6 7)
where an interleave of 2 would put one sector in between
(0 4 1 5 2 6 3 7) or three (0 3 6 1 4 7 2 5).
The idea was
that the typical read/write pattern would be to access the sectors
in consecutive order; but if your machine was too slow, the logical next
sector would pass under the head before it was ready to do something with
it, and then you'd have to wait for the entire rest of the disk to
spin by to get the sector you wanted,
giving a worst case of one revolution
per sector, which would be horrible.
So, you interleave the sectors, so that sectors you don't (yet) care about
are going under the head while the computer is busy calculating the
checksum for the next sector (or whatever), and thus, reduce the number
of revolutions required to read the sectors.
You might have wanted to low level format your hard drive after a cpu upgrade
so to re-optimize the interleave. You might want to reformat a floppy
with a different interleave to use it for an atypical purpose.
Of course, these days, this is nearly a totally obsolete concept.
It was much more relevant when the low level disk bit banging was
implemented in software.
These days, interleave is typically set in stone and optimized
for the disk controller attached to the drive, which usually includes
track buffers that can read in the entire track in one pass anyway, so sequential sectors
on modern disks are most probably contiguous.
Also, most Real operating systems implement read ahead, which
tries to predict the next needed sector, and buffer it.
Other semi-obsolete (and more advanced) disk optimization concepts include
head optimization and spindle optimization, which were implemented in
early versions of BSD UNIX. Today's closest equivalent would include
stuff like the elevator algorithm, which is still relevant because
it is a big picture kind of thing.
This was brought to you by the Save Our Archaic
Technical Terms Society.