high speed drives

[IT Manager]

In article <1351 at icdi10.UUCP>, fr at icdi10.UUCP (Fred Rump from home) writes:
> The Micropolis 1684 SCSI drives promise something like 3.9 MS 'effective' 
> access time with something like a 4.0 MB/sec transfer rate.
> 
> Sounds great.
> 
> Anybody use these yet? And how fast are they really? 
> What controllers have you used?
> I'm particularly interested for functionality under SCO UNIX or Xenix.
> 

We installed a Micropolis 1684 (5.25", 1/2 height, 323MB available to Xenix)
on a Xenix system a few weeks ago.  The host adapter is an Adaptec 1542B
(it has worked with the 1542A as well).  I have been satisfied with its
performance, but we ran into a few incompatibilities.  For instance, this
drive works fine right now when teamed with a Miniscribe 8380S drive.  But,
when teemed with either a Syquest SQ555 or Miniscribe 3180S on the same
SCSI bus, we would get system lockups.  Also, when the 1684 and 3180S were
on the same bus, it seemed like there were times when the 1684 would power
off, and then back on again.

This isn't new to me, as we had other compatibility problems with SCSI
drives in the past.  It's always been pretty easy to just use two host
adapters and divide the drives into two groups that work well together.

As far as performance goes, here's some numbers:

from "time dd if=/dev/XXXX of=/dev/null bs=100k count=10"

for the Miniscribe 3180S
10+0 records in
10+0 records out

real        17.7
user         0.0
sys          0.4

for the Miniscribe 8380S
10+0 records in
10+0 records out

real        17.9
user         0.0
sys          0.5

for the Micropolis 1684
10+0 records in
10+0 records out

real         3.9
user         0.0
sys          0.6

On the surface (and definitely under this installation) the Micropolis is
much faster.  These drives are are installed in their "default" state, so
there has been no performance tuning.  These drives will be involved in an
upgrade soon, and before we install the OS, I'll use Roy Neese's SCSICNTL
program to try to do some fine tuning on the drives.

BTW, using the built-in diagnostics on the 386 these drives are installed
on shows that the Micropolis and Miniscribe 8380S (which are mapped to
drive C: and drive D: by the Adpatec BIOS) have similar performance (I don't
remember the exact numbers, but they were in the 900K/sec neighborhood),
with the Micropolis having a much faster seek time (in the 4ms range, 
compared to about 7.5 for the Miniscribe).  This might explain the difference
in performance as seen under Xenix.

I really like the price, performance, and reliability of SCSI drives, but the
incompatibilities I keep seeing make me wonder some time just how
"standard" SCSI really is.
------------- 
James B. O'Connor			jim at tiamat.fsc.com
Ahlstrom Filtration, Inc.		615/821-4022 x. 651