*BSD News Article 79318


Return to BSD News archive

Newsgroups: comp.unix.bsd.freebsd.misc,comp.unix.bsd.bsdi.misc
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!news.wildstar.net!cancer.vividnet.com!hunter.premier.net!news1.erols.com!howland.erols.net!psinntp!psinntp!cmcl2!rockyd!dnn.rockefeller.edu!dan
From: dan@dnn.rockefeller.edu (Dan Ts'o)
Subject: Re: Why one should buy parity memory for reliability?
X-Nntp-Posting-Host: dnn.rockefeller.edu
Message-ID: <DyDE1r.B9J@rockyd.rockefeller.edu>
Followup-To: comp.unix.bsd.freebsd.misc,comp.unix.bsd.bsdi.misc
Sender: notes@rockyd.rockefeller.edu (News Administrator)
Organization: Rockefeller University
X-Newsreader: TIN [version 1.2 PL2]
References: <32485B0D.41C6@austin.ibm.com> <52b1lk$ijl@hamilton.maths.tcd.ie> <52bkb4$h0i@shellx.best.com>
Date: Fri, 27 Sep 1996 02:39:27 GMT
Lines: 21
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:28089 comp.unix.bsd.bsdi.misc:4993

In article <52bkb4$h0i@shellx.best.com> you wrote:
: In article <52b1lk$ijl@hamilton.maths.tcd.ie>,
: David Malone <dwmalone@maths.tcd.ie> wrote:
: >I guess the big thing is that you will get definite failures, as
: >opposed to the machine running funny from time to time. Its always
: >such a pain if something works properly half the time. When your
: >machine crashes with a parity error at location blah, you know your
: >simm is probably at the end of its days.
: Right.  The essential question is how much is your (or the person
: who has to care that the thing keeps running) time worth?  With parity
: the system halts with an NMI message of some sort.  Without it, it
: crashes, usually sporadically, often with different symptoms.

	Actually you are the describing the best case failure scenerio, that
the machine crashes or acts funny.
	The worse case failure scenerio is when it doesn't crash or act funny.
It just corrupts your data or results silently. Suppose you have a non-parity
simm in high memory (e.g. 64Mb). And that simm rarely gets used because the
machine is so lightly loaded *except* for that once-per-month payroll run,
which is a big program that is about the only program that uses that simm.
You may well get a corrupted, but potentially otherwise silent payroll run.