Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!munnari.oz.au!metro!dmssyd.syd.dms.CSIRO.AU!ditsydh.syd.dit.CSIRO.AU!evans
From: evans@syd.dit.CSIRO.AU (Bruce.Evans)
Newsgroups: comp.os.minix
Subject: Re:  fdisk, head, cylinders and blocks
Message-ID: <1991May25.035438.5121@syd.dit.CSIRO.AU>
Date: 25 May 91 03:54:38 GMT
References: <53806@nigel.ee.udel.edu> <1991May22.050418.189@skypod.guild.org> <4463@stl.stc.co.uk>
Organization: CSIRO Division of Info Tech, Sydney, Australia
Lines: 20

In article <4463@stl.stc.co.uk> agm@daphne.stl.stc.co.uk (Andrew G. Minter) writes:
>There are however some implicit assumtions about disk geometry in the
>FS.  It's in read.c if I remember correctly and it's to do with doing
>read-ahead.  Minix is supposed to read ahead the rest of the track
>following a read to increase speed.  The trouble is, if the FS has a
>totally wrong idea about your geometry this can actually slow the file
>system down.  Another problem here is that IDE drives often lie about
>their real geometry and pretend to have a different layout.  They also
>tend to do track buffering in hardware.
>
>My inclination is to totally disable the read-ahead stuff.

This is not a good idea. FS guesses 17 for the number of sectors per track.
Disabling read-ahead would have a side effect equivalent to guessing 1
sector per track. Actually it doesn't matter much for hard disks because
they either have decent buffering in the hardware or interleave is used to
slow them down enough for the software to keep up. Then any large amount
of read-ahead works well (except for random i/o).
-- 
Bruce Evans		evans@syd.dit.csiro.au