home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!haven.umd.edu!mimsy!prometheus!media!irscscm!bogart!mfaurot
- From: mfaurot@bogart.uucp (Michael Faurot)
- Newsgroups: comp.os.coherent
- Subject: Re: Hard drive performance
- Message-ID: <1993Jan19.183259.4738@bogart.uucp>
- Date: 19 Jan 93 18:32:59 GMT
- References: <1j2fmuINNpen@menudo.uh.edu>
- Organization: ism:c:m/xo
- Lines: 30
- X-Newsreader: Tin 1.1 PL4
-
- urban@euler.cl.uh.edu (MARCUS A URBAN) writes:
- : Is anyone out there getting decent hard disk IO performance under Coherent?
-
- On my ST-506/RLL drives I'm getting anywhere between 100K-209K of
- throughput depending on the particular partition being tested.
-
- Judging from the numbers you listed in your original post (29K-69K)
- you'll probably need to do some experimenting with the "n" and "m"
- values used by mkfs when it creates filesystems. When I first
- installed Coherent and mkfs created the filesystem with whatever it
- uses for defaults I had the same lousy throughput as you describe.
- After spending a day doing some experimenting with different values
- and Iozone to provide the benchmarking I was able to get the figures I
- mentioned above.
-
- What you should do to experiment is use an idle partition if at all
- possible and try different values until you start getting better
- figures from Iozone. Be sure to do a back-up though, even if working
- on another partition, because it's very easy to misstype the name of a
- partition and mkfs doesn't ask twice before it blows away everything.
-
- If anyone reading this would like a nice project, how about a utility
- that could figure out the optimal values for mkfs's "n" and "m"
- parameters. It would sure beat trial and error.
- --
-
- +--------------------+-------------------------------------------------------+
- | Michael Faurot | Domain: mfaurot@bogart.UUCP |
- | ------- ------ | UUCP: ...{irscscm|mimsy}!bogart!mfaurot |
- +--------------------+-------------------------------------------------------+
-