My Original Post:
| Q. Could someone please help me verify which is best for my application.
| Q. What happens if i use the wrong format.dat with a disk ??
| Q. Can I use the 2nd entry and get 2.4gb instead of 2.1gb from
| the same disk ??
|
| ----------my-vendor-------------
| disk type="Seagate ST-12400N 2.1GB"\
| :ctlr=SCSI\
| :ncyl=2709 :acyl=2 :pcyl=2711 :nhead=19 :nsect=80\
| :rpm=5400 :bpt=40960:
|
| --------circulating with sun-managers--------
|
| disk_type = "Seagate ST12400N 2.4GB" \
| : ctlr = MD21 : fmt_time = 4 \
| : cache = 0x0 : trks_zone = 19 : asect = 1 : atrks = 0 \
| : ncyl = 2668 : acyl = 2 : pcyl = 2621 : nhead = 19 : nsect = 82 \
| : rpm = 5411 : bpt = 41984
| ---------------------------------
SUMMARY: I got two very good and detailed responses to my post. Going thru
them it was clear that Number of data cylinders do not matter as SunOS 4.1X
(not sure about 5.X ) uses blocks not cylinders for it's SCSI disk organisation.
If you have pcyl, acyl, nhead, nsect, rpm, bpt(not sure) values which
normally comes with the disks. Use ncyl=pcyl-acyl and you have your format.dat.
Increasing the "ncyl" is not going to get you any extra space on the same disk,
Best value for ncyl is (pcyl minus acyl).
I used the second entry and it worked fine with me ( I did not try the first
one as the values -rpm,pcyl did not match with the disk-specs which came with
the disk ).
One other very important thing is to never mess with the original formatted
disk from the manufacturer, simply repartition the disk according to your
need. Re-formatting should be avoided as far as possible.
Anyway read thru these reponses and everything should be crystal clear ...
--------------------------------------------
perryh@pluto.rain.com (Perry Hutchison) Wrote :-
When dealing with SCSI disks, the C/H/S geometries are almost entirely
fictitious as the drive is addressable only by block number. It is
conceiveable that specifying the actual internal geometry would result
in somewhat better performance -- the 4.2 filesystem was originally
designed to optimize various latencies back in the SMD days when the
system had to know the actual geometry anyway -- but I am not aware of
anyone having tested this recently. It seems likely that the internal
buffering and zoned recording techniques used in modern SCSI drives
would hide most of the effect of these optimizations.
I suppose you could try both, and see if there is any difference in
performance. If not, and if the second entry works, you might as well
use it and increase your available storage.
-------------------------------------
Kevin Martinez <lps@rahul.net> Wrote:-
The best format.dat entry will be the one that matches the total number
of LBAs (512 byte sectors available for the end user). The documentation
with the drive or a call to Seagate should confirm the actual LBA count.
It is also available from the drive if you have the means to issue a SCSI
Inquiry (12h) or a Read Capacity (25h) command.
The format.dat entries ncyl, nhead and nsect multiplied together should
give a number less than or equal to the actual number of LBAs. The ideal
case is if they are equal. Less than means you are not using the full
capacity of the drive. If it is greater than the number of LBA available
and you try to make a filesystem that uses these non-existant LBAs, count
on major problems with your host.
If you can issue SCSI commands from your host, the results can be used
with the unix "factor" command to make your own format.dat entries.
---------------------------------------
perryh@pluto.rain.com (Perry Hutchison)
Kevin Martinez <lps@rahul.net>
Lloyd C. Cha <lccha@nb.rockwell.com>
slee.ATC@madge.com (Steve Lee Info:SJ)
Thanks a lot for your responses,
..Vikram
-- Vikram Dutta Email: vikram@snt.com Software & Technologies Inc. Fax: (408)496-0932
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:00 CDT