Return to BSD News archive
Newsgroups: comp.os.386bsd.questions
Path: sserve!newshost.anu.edu.au!munnari.oz.au!bunyip.cc.uq.oz.au!news.qut.edu.au!orac.crissp.qut.edu.au!reilly
From: reilly@orac.crissp.qut.edu.au (Andrew Reilly)
Subject: Re: DISK LIGHT HANG !#@*&!!!
Message-ID: <1993Nov8.015518.23306@news.qut.edu.au>
Sender: news@news.qut.edu.au (USENET News System)
Reply-To: A.Reilly@qut.edu.au
Organization: Signal Processing Research Centre, QUT, AUS
References: <1993Nov3.190751.17591@sifon.cc.mcgill.ca>
Date: Mon, 8 Nov 93 01:55:18 GMT
Lines: 47
In article <1993Nov3.190751.17591@sifon.cc.mcgill.ca>, fox@cs.mcgill.ca (Colin BRADLEY) writes:
> I have the disk light hang problem. I know this has been
> posted about before, but were any viable solutions
> presented?
>
> Oh, I am running NetBSD-0.9...
I am too.
I know this doesn't look like the source of the problem, but
perhaps the VM implementation has something to do with it:
> swap 32 meg swap 20 meg
I would get the HD light crash 2 or more times a day when I
only had 12M swap configured. Always after a bout of really
hard paging. I have since re-configured to 32M swap, and
don't have the problem at all. Of course disk usage is
reduced, because to get 32M for swap I had to move the /usr
partition off to my NFS host.
Incidentally, I'm using a 33MHz 486 box with one of those
all in one: floppy, ports, IDE interface cards connected to
a Connor 80M drive. I have a WD8013 ethernet card that
works really well with David Greenman's ed driver. Here's
my question:
This is in many respects a fairly quick machine (not
blazing, but OK.) However the disk performance seems to be
disproportionately slow, compared to my experiances of Unix
on workstations (my usual environment.) Does anyone have an
opinion on the cause? Is it the AT bus? Presumabably
anything would work faster with an EISA or VL bus, but would
an EISA or VL bus IDE interface work as well (fast) as a
DMA-based SCSI interface? What about a DMA-SCSI interface
on the AT bus?
The way I notice the performance "problem" is that
everything else stops while disk IO is happening. That just
doesn't occur on the workstations that I'm used to, all of
which use DMA to access SCSI drives. So where are the CPU
cycles going with IDE?
-- Andrew Reilly -- | A.Reilly@qut.edu.au |
Signal Processing Research Centre |
QUT, GPO 2434, Brisbane 4001, Australia. |
phone: +61 7 864 1400 | fax: +61 7 864 1516 |