Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!mcsun!ukc!reading!minster!martin
From: martin@minster.york.ac.uk
Newsgroups: comp.arch
Subject: Memory utilization & inter-process contention
Message-ID: <619868166.27927@minster.york.ac.uk>
Date: 23 Aug 89 09:36:06 GMT
Organization: Department of Computer Science, University of York, England
Lines: 25

This seems to be one area in which ``progress'' in the last few years
(especially on Unix systems) appears to have been going backwards!!

Several systems in the past have had very good (although still very simple)
paging systems which minimise the kind of effects that have been mentioned.
A good example that springs to mind was the Edinburgh EMAS system.

Given existing work, there is no excuse for a system to EVER thrash.
If a process gets too big, it might slow down, but it should not affect
other processes (to first order, anyway).

Recently too much effort has (I my opinion :-) been put into ``feature-full''
paging systems, and not enough into paging systems that work. Dare I say it,
Berkeley and SunOS leap to mind here.

Martin
usenet: ...!mcvax!ukc!minster!martin
JANET:  martin@uk.ac.york.minster
surface:
	Martin C. Atkins
	Dept. of Computer Science
	University of York
	Heslington
	York YO1 5DD
	ENGLAND