home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.hp
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!rpi!psinntp!psinntp!polari!lampi@polari.online.com
- From: lampi@polari.online.com
- Subject: Re: Third party disk drives on HP-UX
- Message-ID: <1992Sep4.201300.9158@polari>
- Summary: Come on in---the water's fine! :-)
- Keywords: HP-UX disk drive 3rd third party
- Organization: R Squared, Redmond, WA (206) 883-3116
- References: <1992Aug29.191511.29259@riacs.edu> <trn.715289167@warper.jhuapl.edu>
- Date: Fri, 4 Sep 1992 20:13:00 GMT
- Lines: 91
-
- In article <trn.715289167@warper.jhuapl.edu> trn@warper.jhuapl.edu (Tony Nardo) writes:
- >okuyama@pioneer.arc.nasa.gov (Darin Okuyama YR) writes:
- >
- >>Has anyone successfully installed a thrid-party SCSI disk
- >>dirve on a Snake (HP 9000/700)?
- >
- >I installed a pair of disks from ADS (contained 2 Seagate ST41600N Elite
- >drives, each with ~1.3 Gbytes storage). So far they seem to work.
- >
- >
-
- Being a third party, yes :-)
-
- >>If you have, then:
- >
- >> (1) Was it relatively straightforward?
- >
- >Moderately. The hardest part was researching the drive's operating
- >parameters (i.e. # of sectors, # of tracks, # of cylinders). Beyond that,
- >SAM gave me the info I needed to add support for the drive. SAM will tell
- >you which files to modify so that you may add a new "Disk drive model" to
- >the system.
- >
- Very straightforward, but with one caveat (see below)
-
- >> (2) What problems did you encounter?
- >
- >I lost roughly 7% of the usable space on my disk due to the assumption by
- >HP that all drives will have 1024-byte sectors. They provide a workaround
- >solution for those who have 512-byte sectors so that you may still use the
- >drive. However, a disk with an odd number of 512-byte tracks will lose one
- >track under the workaround.
- >
- Make that 1 block (512 bytes).
-
- >I did not have access to the rotational speed of my drive, but 3600 seemed to
- >be a fairly common number. I used that, and it worked. Whether I guessed
- >correctly or whether the HP driver no longer uses this information, I cannot
- >say.
- >
- The value for rpm seems to be ignored.
-
- >> (3) What advice can you give the Net?
- >
- >1) Have your drive's documentation available to determine the # of tracks,
- > sectors, & cylinders. If you don't have this documentation, but you have
- > a similar model drive on a Sun, you can obtain this information with the
- > "dkinfo" command. Other flavors of Unix may offer similar ways to
- > inspect disk information on other systems.
-
- This information is not really necessary. (see below)
- >
- >2) Obtain the rotational speed of the drive. (I was willing to take a
- > chance on 3600, lacking the proper number, but this may not be a good
- > idea.)
-
- This is not really necessary. (see below)
- >
- >3) Use SAM to add the disk drive. The files you need to modify are well
- > documented, and SAM tells you which files you'll need to change. (Those
- > files, BTW, are /usr/sam/lib/disk_aliases and /etc/disktab.
- >
- Or you can do it by hand & just modify /etc/disktab.
-
- >Good luck!
- >--
- >Tony Nardo, INET: trn@warper.jhuapl.edu, trn@aplcomm.jhuapl.edu
- > Johns Hopkins Univ./APL UUCP: {backbone!}mimsy!aplcomm!trn
-
- (Below)
- I've just run through some 300+ benchmark runs with different disk drives
- (total of 6 drives) and different disktab entries. Here are the results
- (which I presented at a Hardware & Peripherals SIG session at InterWorks):
-
- 1. Optimal disk geometries have no obvious relation to the physical
- geometry of any given disk drive.
- 2. Different disk geometries for a given disk drive can cause that disk
- drive to perform as far as 50% slower than optimal on sequential I/O.
- 3. Different disk geometries for a given disk drive can cause that disk
- drive to have writes perform much faster, the same as, or much slower
- than reads during sequential I/O.
- 4. The values specified for rpm, sector size and number of cylinders (rm#,
- se# and nc#) are ignored, at least in 8.0x.
- 5. Selecting the best geometry for your application may require the
- running of a series of benchmarks on your system.
-
- --
- -----
- +-+ Michael Lampi lampi@polari.online.com
- |R| R Squared 16398 NE 85th St., Suite 101, Redmond, WA 98052
- +-+ (206) 883-3116 fax (206) 883-2676
-