Some (Unix PC) HD facts and observations.

Augustine Cano afc at shibaya.lonestar.org
Mon Dec 24 09:22:51 AEST 1990 

I finally have HD space to spare after I installed my new Priam 519 from
the group buy.  For those who don't know, the 519 is a (faster) equivalent
of the Maxtor 2190 (1224 cylinders, 15 heads, 22 ms avg access time.)

Both the P5.1 upgrade and a WD2010 are needed to take full advantage of it.
Some observations follow:

Power requirements:

The label on the 519 lists the following:  +5V @ 1.7A, +12V @2.6A.
The old ST4096 (1024 cylinders, 9 heads) had this: +5V @ 1.5A, +12V @4.00.
The strange thing is that even in the most active times, the old drive
did not affect the monitor; the new one, when doing doing lots of seeks,
makes the display wiggle.  It must really be sucking power...  Did anybody
else notice this?

Access times:

Timing the random seek test (1000 random seeks) on the diagnostics disk
provided a surprisingly consistent set of numbers (in s): 25.27, 25.03,
25.00, 25.20, 25.24, 25.12, 24.96, 25.12, 25.00, 25.42.  For all practical
purposes, 25 ms real access time.  I speculate that the 3 ms difference
between this in the official access time is caused by the slower step
rate that is the default in the unix pc.  More on step rates later.

Bad blocks:

I wiped out the BBT repeatedly and left the bad block scan test running
overnight a few times.  If the bad blocks on the manufacturer defect list
were not entered, the test program found only a small minority of the
flaws.  Since I entered the bad blocks manually, not a single HDERR line
has appeared on /usr/adm/unix.log.

HDERR crashes:

Not a single one.  It turned out that the culprit was obviously the old
ST4096.  The Seagate drive had only 6 entries in the manufacturer bad block
list, so it is likely that the testing done to find those defects was MUCH
LESS stringent.  When I ran the bad block scan on the Seagate drive without
entering the BBT manually, 5 defects were found on the first run, the sixth
one took much longer but it was eventually detected too.  I wonder if all
the HDERR crashes and entries in /usr/adm/unix.log, that appeared to occur
all over the disk, were not really random or caused by power supply
fluctuations but were in fact caused by marginal coating on the platters.
A more stringent test by the manufacturer would have probably found so
many defects as to make the drive a complete reject.

Step rate:

One of the first things I did after I had everything re-installed was to
try the "fast" driver written by Peter Fales (thanks Peter!).  Then I ran
the disk exerciser written by Lenny Tropiano (thanks Lenny!).  The results
are included below.  Judge for yourself.
The subjective feel is much crisper.  I can tell the difference.

==========================================================================

UNIXPC Disk Exerciser for device: /dev/rfp000

Disk name: WINCHE   Cylinders: 1224   Heads: 15
Step rate:  0

Max Seek Distance = 150404608
** Converging Seeks ... 
iter[0]: 2426
iter[1]: 2456
iter[2]: 2448
iter[3]: 2447
iter[4]: 2438

Average seek time: 2443

** Sequential Seeks ... 
iter[0]: 1093
iter[1]: 1768
iter[2]: 1107
iter[3]: 1096
iter[4]: 1143

Average seek time: 1241

--------------------------------------------------------------------------

UNIXPC Disk Exerciser for device: /dev/rfp000

Disk name: WINCHE   Cylinders: 1224   Heads: 15
Step rate: 14

Max Seek Distance = 150404608
** Converging Seeks ... 
iter[0]: 2036
iter[1]: 1981
iter[2]: 1982
iter[3]: 1987
iter[4]: 1997

Average seek time: 1996

** Sequential Seeks ... 
iter[0]: 1098
iter[1]: 1099
iter[2]: 1095
iter[3]: 1094
iter[4]: 1115

Average seek time: 1100

==========================================================================

UNIXPC Disk Exerciser for device: /dev/rfp000

Disk name: WINCHE   Cylinders: 1224   Heads: 15
Step rate:  0

Max Seek Distance = 150404608
** Long Seeks ... 
iter[0]: 5
iter[1]: 8
iter[2]: 8
iter[3]: 7		# NOTE: the ... replace identical lines
...
iter[19]: 7
iter[20]: 8
...
iter[23]: 8
iter[24]: 9
iter[25]: 8
...
iter[31]: 8
iter[32]: 9
iter[33]: 10
iter[34]: 30
iter[35]: 8
...
iter[49]: 8

Average seek time: 8

** Random Seeks ... 
iter[0]: 5 (Cyl=109, Head=2)
iter[1]: 3 (Cyl=365, Head=7)
iter[2]: 4 (Cyl=200, Head=6)
iter[3]: 5 (Cyl=1169, Head=9)
iter[4]: 5 (Cyl=196, Head=12)
iter[5]: 5 (Cyl=439, Head=15)
iter[6]: 6 (Cyl=1194, Head=14)
iter[7]: 7 (Cyl=817, Head=9)
iter[8]: 5 (Cyl=410, Head=4)
iter[9]: 5 (Cyl=538, Head=15)
iter[10]: 4 (Cyl=13, Head=5)
iter[11]: 3 (Cyl=456, Head=6)
iter[12]: 4 (Cyl=195, Head=14)
iter[13]: 5 (Cyl=1011, Head=9)
iter[14]: 5 (Cyl=770, Head=4)
iter[15]: 3 (Cyl=91, Head=2)
iter[16]: 4 (Cyl=531, Head=14)
iter[17]: 4 (Cyl=35, Head=7)
iter[18]: 3 (Cyl=348, Head=15)
iter[19]: 3 (Cyl=351, Head=1)
iter[20]: 3 (Cyl=933, Head=3)
iter[21]: 5 (Cyl=934, Head=9)
iter[22]: 6 (Cyl=795, Head=12)
iter[23]: 5 (Cyl=439, Head=10)
iter[24]: 4 (Cyl=838, Head=1)
iter[25]: 3 (Cyl=542, Head=10)
iter[26]: 5 (Cyl=627, Head=12)
iter[27]: 6 (Cyl=899, Head=10)
iter[28]: 6 (Cyl=825, Head=13)
iter[29]: 10 (Cyl=1022, Head=1)
iter[30]: 8 (Cyl=532, Head=3)
iter[31]: 3 (Cyl=45, Head=7)
iter[32]: 3 (Cyl=251, Head=11)
iter[33]: 4 (Cyl=497, Head=4)
iter[34]: 4 (Cyl=154, Head=12)
iter[35]: 6 (Cyl=1164, Head=15)
iter[36]: 6 (Cyl=910, Head=3)
iter[37]: 4 (Cyl=171, Head=4)
iter[38]: 3 (Cyl=1132, Head=1)
iter[39]: 4 (Cyl=371, Head=13)
iter[40]: 5 (Cyl=432, Head=14)
iter[41]: 4 (Cyl=405, Head=13)
iter[42]: 4 (Cyl=519, Head=8)
iter[43]: 5 (Cyl=689, Head=12)
iter[44]: 6 (Cyl=1035, Head=8)
iter[45]: 5 (Cyl=595, Head=2)
iter[46]: 4 (Cyl=874, Head=1)
iter[47]: 3 (Cyl=560, Head=2)
iter[48]: 4 (Cyl=749, Head=1)
iter[49]: 5 (Cyl=1183, Head=15)

Average seek time: 4

--------------------------------------------------------------------------

UNIXPC Disk Exerciser for device: /dev/rfp000

Disk name: WINCHE   Cylinders: 1224   Heads: 15
Step rate: 14

Max Seek Distance = 150404608
** Long Seeks ... 
iter[0]: 5
iter[1]: 6		# NOTE: the ... replace identical lines
...
iter[49]: 6	<-------+
			|-- displayed # always rounded up?
Average seek time: 5 <--+

** Random Seeks ... 
iter[0]: 4 (Cyl=697, Head=1)
iter[1]: 3 (Cyl=799, Head=6)
iter[2]: 4 (Cyl=903, Head=1)
iter[3]: 3 (Cyl=173, Head=14)
iter[4]: 4 (Cyl=271, Head=6)
iter[5]: 4 (Cyl=494, Head=11)
iter[6]: 4 (Cyl=1084, Head=13)
iter[7]: 5 (Cyl=1132, Head=14)
iter[8]: 4 (Cyl=774, Head=12)
iter[9]: 2 (Cyl=574, Head=1)
iter[10]: 3 (Cyl=1147, Head=5)
iter[11]: 3 (Cyl=1181, Head=6)
iter[12]: 3 (Cyl=894, Head=7)
iter[13]: 4 (Cyl=864, Head=14)
iter[14]: 4 (Cyl=479, Head=4)
iter[15]: 4 (Cyl=711, Head=14)
iter[16]: 3 (Cyl=287, Head=10)
iter[17]: 3 (Cyl=849, Head=5)
iter[18]: 3 (Cyl=1202, Head=7)
iter[19]: 2 (Cyl=62, Head=11)
iter[20]: 3 (Cyl=650, Head=15)
iter[21]: 4 (Cyl=900, Head=15)
iter[22]: 4 (Cyl=976, Head=2)
iter[23]: 3 (Cyl=483, Head=6)
iter[24]: 3 (Cyl=754, Head=9)
iter[25]: 3 (Cyl=1215, Head=4)
iter[26]: 2 (Cyl=3, Head=13)
iter[27]: 1 (Cyl=80, Head=4)
iter[28]: 2 (Cyl=509, Head=12)
iter[29]: 3 (Cyl=691, Head=9)
iter[30]: 3 (Cyl=535, Head=3)
iter[31]: 3 (Cyl=822, Head=10)
iter[32]: 3 (Cyl=54, Head=15)
iter[33]: 3 (Cyl=1038, Head=14)
iter[34]: 3 (Cyl=581, Head=1)
iter[35]: 2 (Cyl=705, Head=2)
iter[36]: 4 (Cyl=733, Head=15)
iter[37]: 5 (Cyl=1058, Head=9)
iter[38]: 4 (Cyl=858, Head=15)
iter[39]: 5 (Cyl=776, Head=12)
iter[40]: 3 (Cyl=291, Head=9)
iter[41]: 4 (Cyl=1206, Head=12)
iter[42]: 4 (Cyl=583, Head=7)
iter[43]: 3 (Cyl=304, Head=15)
iter[44]: 2 (Cyl=560, Head=1)
iter[45]: 3 (Cyl=794, Head=12)
iter[46]: 3 (Cyl=1025, Head=4)
iter[47]: 3 (Cyl=961, Head=6)
iter[48]: 4 (Cyl=863, Head=14)
iter[49]: 4 (Cyl=589, Head=5)

Average seek time: 3

-- 
Augustine Cano		INTERNET: afc at shibaya.lonestar.org
			UUCP:     ...!{ernest,egsner}!shibaya!afc

More information about the Comp.sys.att mailing list