----------
> From: Russ Poffenberger <[log in to unmask]>
> Subject: Re: [PCBUILD] SCSI Info
> Date: Tuesday, July 07, 1998 10:19 AM
>
> Furthermore, many modern disks use a zone recording method where the
> sectors/track varies from the inside to the ouside of the platters. This
is
> simply because the outer tracks have a greater circumference and can hold
more,
> so the data transfer is higher when reading the outer tracks.
>
Please forgive a little thinking out loud. I'm thinking about the
different factors that could affect disk performance ... I guess mainly I
need an update on the different factors that affect disk performance in
modern disk drives as opposed to the ones I knew about years ago.
Is this really true ... the part about the number of sectors per track
changing? I know that this was true years ago when the recording surface
on the platter was several inches across, but with today's smaller platter
surfaces, I would not think there is enough difference in the length of the
tracks to bother with this. But then, with the higher recording density
possible now, the length of a sector is much smaller and when that is
compared to the length of the track, it might be feasible and effective to
vary the sectors per track. Anyone know the physical length of a sector on
a modern disk drive? -- I would assume that it depends on the recording
density and the spin rate of the drive. Actually, I guess it also depends
on the track position since the platter is a solid body which results in
the length of the surface that passes under the head in a fixed time to be
greater for the outer tracks.
A little background:
You should understand that I don't have lot of experience working with PC
hard drives at this level. It seems to me that, at least on the mainframe
disk drives I worked with a few years ago, that the number of sectors per
track was constant and the difference in the length of the tracks was
accomodated by increasing the length of the space between the sectors. In
the case of a disk where the number of sectors per track is constant for
all tracks, the inner tracks will have a higher transfer rate because the
spaces between the sectors will be smaller. However, you could lose
transfer rate if the controller could not keep up with the spin rate and
you missed a sector, resulting in having to wait for the next sector to
come around again. An attempt was made to reduce the "lost rev" wait time
by interleaving sectors to increase the amount of time the controller had
to get rid of the data before reading in the next "logical" sector. That
is, sectors were not stored in strictly sequential order. A 2-to-1
interleave meant that the sectors were physically recorded in every other
physical sector such that (for example a 7 sector/track disk) the order
would be 0,4,1,5,2,6,3 and it would take two revolutions of the disk to
read the entire track. Years ago, it appears that the performance of disk
controllers and CPUs eliminated the need for inter-leaving, but now that
the spin rates of drives are increasing all the time, I wonder if this is
still true.
Does anyone see anything wrong with this analysis? Assume a platter
diameter of 3" with a recordable surface area 1" across (which I seem to
remember hearing as being the spec for a "modern" hard drive). Then the
inner track is about 6.25" long and the outer track is about 9.4" long or a
50% difference. With that much change in track length (at first thought, I
did not think it would be that great), I guess it could make sense to
change the sectors per track, but I wonder if the performance of the
controller would be affected by having to keep track of the different
recording zones. Anyone have any information on this part of the problem?
This is all a bit rambling, but it points out some of the different factors
that must be considered when designing a high-performance disk drive. I
know there that years ago there was a lot of trade-off between these
factors that affected disk performance. I'm wondering how these trade-offs
are being managed today.
|
