🚀 go-pugleaf

In article <1992Dec18.145311.27416@sequent.com>, cyf@sequent.com (Cyrus Foughty) writes:
> I have heard two theories about one element that affects disk transfer
> rate.

> Theory 1: 
> 
> A higher transfer rate will result when using the outer tracks.

This is true only if the drive uses constant density recording.  It is quite 
common, especially on older disks to lay down the the same amount of bits 
regardless of radii, thus the xfer speed would be exactly the same.
 
> Theory 2:
> 
> A higher transfer rate will result when using the inner tracks.

Never true (to my knowledge).

The important thing to remember is that the inside of the platter and
the outside of the platter take the same amount of time to make one
revolution.  What can change is the number of bits that are recorded on
a track.
  

cyf@sequent.com (Cyrus Foughty) writes:

>I have heard two theories about one element that affects disk transfer
>rate. Help me disect these and find which is true and which is false.
>Of course I feel that there will be some cases where both are true and
>both are false. So, based on that I will tell you that I am dealing 
>with the HP97560 SCSI-2 disk drive.

>Theory 1: 

>A higher transfer rate will result when using the outer tracks. "outer"
>is defined here as "tracks that are the greatest distance from the 
>spindle". I.E., track numbering starting at zero.

>Theory 2:

>A higher transfer rate will result when using the inner tracks.
>"inner" is defined here as "tracks that are the closest to the
>spindle". I.E., track numbering starting at maximum track number and 
>decreasing.

>                       Stuff Deleted

Two months ago I didn't know about zones on disks that take advantage
of the greater circumference of the outer zones.  It not only increases
capacity of a disk but performance as well.  I was using a Seagate Elite
disk to measure SCSI bus performance and found that if I moved to a 
different but equivalent system, I often couldn't duplicate the best
performance I had achieve elsewhere.  I used

	 timex dd if=/dev/rdsk/c0t0d0s9 of=/dev/null bs=16384 count=1024

and got a bandwidth of about 2.7 MB/second but if I used slice 0, I 
would get about 3.6 MB/second.  Evidently, the way we assign slices,
s0 starts on the outer track and works in.  Slice 9 is in near the
hub.  When the performance seem to be slice-dependent, we went to the
trouble of reading the disk specification.  Seagate calls the technique
"Zone Bit Recording" and trademarks "ZBRtm". The disk is divided into
14 zone where the sectors per track range from 85 on the outside to
61 on the inside.  Zone frequency and bytes/track correspond similarily.

A Seagate document says "Variable Drive Serial Transfer Rate (26 MHz - 
36 MHz) is transparent to host due to buffered interface."  The 
variable transfer rate was not transparent in my test.  We may not 
have tuned the buffer interface.

I don't know about the HP97560, specifically, but I have heard that
some HP disks do something like this but with just three zones, of
76, 88, and 96 sectors per track.

Not a spokesperson for anyone.

-- 
======================================================================
= Martin D. Sandman                                     .........    =
= NCR Torrey Pines                                      ..x......    = 
= 11010 Torreyana Road    martins@TorreyPinesCA.NCR.com ....x....    =
= San Diego, CA 92121     Phone: 619-597-3562           .xx..xxx.    =
= U. S. A.                FAX:   619-597-3501           .........ants=
======================================================================

martins@TorreyPinesCA.ncr.com (Martin Sandman 3562) writes:
>Two months ago I didn't know about zones on disks that take advantage
>of the greater circumference of the outer zones.  It not only increases
>capacity of a disk but performance as well.  

      Yes, provided you have some kind of high-level drive interface where
true drive layout is hidden from the software, such as IDE or SCSI, this
is a good way to improve drive density.  It's tougher if the operating system's
driver has to deal with all these zones.

      It does create problems for some disk-optimization software,
though, where attempts are made to position files across track boundaries
for optimal reading speed.  Any info on how programs such as the Norton
Utilities deal with this issue?

      There's also the issue that disk drivers which reorder requests
to minimize head movement may not get the results expected.  But then,
with today's physically small drives, head movement time is comparable to
rotational latency, and the old emphasis on minimizing head movement
no longer applies.

					John Nagle