*BSD News Article 32524


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!gatech!swrinde!pipex!sunic!trane.uninett.no!eunet.no!nuug!EU.net!Germany.EU.net!news
From: bs@Germany.EU.net (Bernard Steiner)
Newsgroups: comp.os.386bsd.questions
Subject: scsi fs tuning
Date: 7 Jul 1994 16:34:26 +0200
Organization: EUnet Deutschland GmbH, Dortmund, Germany
Lines: 26
Distribution: world
Message-ID: <2vh3pi$lq2@Germany.EU.net>
NNTP-Posting-Host: qwerty.germany.eu.net


Hi there,

I seem to have a slight problem with tuning my scsi disk filesystem:

i/o to the raw device reads about 2MB/sec with 64k blocks (10 Meg to
/dev/null), but reading 10 Meg from a file on a 512/4096 filesystem on the
disk gives about the same silly results as reading 4k blocks off the raw
device.

Well, dumpfs told me the rotational delay was assumed to be 4 ms.
After buildig a new filesystem with rot-delay set to 0, I can now get
1MB/sec on continuous read from a 10 Meg file in a 512/4096 filesystem.

Instead of fiddling with obscure parameters - is there any program that tells
me what to do about these parameters (like headswitch tim, track-to-track
seek time, interleave, sector skew, etc.) ?

Or is there a recommended step-by-step refinement towards an optimal set of
parameters ?

And how do I go about disabling the cache on the disk without using a
soldering iron ?

Thanks,
	Bernard