Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!decvax!tektronix!hplabs!hao!seismo!columbia!garfield!kearns
From: kearns@garfield.columbia.edu (Steve Kearns)
Newsgroups: net.micro.mac
Subject: Re: Delphi Mac Digest V2 #11
Message-ID: <1428@garfield.columbia.edu>
Date: Mon, 31-Mar-86 14:22:34 EST
Article-I.D.: garfield.1428
Posted: Mon Mar 31 14:22:34 1986
Date-Received: Thu, 3-Apr-86 04:53:31 EST
References: <4608@topaz.RUTGERS.EDU> <780@wang.UUCP> <2495@sdcc6.UUCP>
Organization: Columbia University CS Department
Lines: 23
Summary: remember what disk caches do...

> > Subject: Hard Disk Benchmark
> > Date: 21-MAR 05:48 Hardware & Peripherals
> >  
> > [summary of benchmark procedure
> As a Mac Plus owner I find these results surprising since I have noticed
> considerably superior performance of my machine to that to non-plus Macs
> I have used.  In particular, I have found the disk cache to be quite
> useful when running large applications (e.g. Excel, MacDraw) which
> normally do a fair amount of segment swapping.  I suspect that the
> structure of the disk cache is such that it does not help much with the
> type of disk access that this benchmark performs.
> 

Disk caches work by remembering blocks of memory that have been previously read
in so that if they are referenced again they can be fetched from ram instead
of the disk.  Thus, unless the application (or benchmark) references the same
data within a reasonable amount of time, the disk cache won't help.  This 
might explain why the benchmark didn't show a performance increase.  Note that
a disk cache is like a ram disk that is loaded "as you go along" based on 
what you loaded so far.  
(For a more comprehensive discussion of disk caches see any good operating
systems book under the topic of "page replacement strategies". )
-steve