Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!caip!clyde!burl!ulysses!mhuxr!mhuxt!houxm!ihnp4!inuxc!pur-ee!uiucdcs!ccvaxa!aglew
From: aglew@ccvaxa.UUCP
Newsgroups: net.arch
Subject: Re: paging
Message-ID: <5100132@ccvaxa>
Date: Sat, 13-Sep-86 18:49:00 EDT
Article-I.D.: ccvaxa.5100132
Posted: Sat Sep 13 18:49:00 1986
Date-Received: Sun, 14-Sep-86 20:12:34 EDT
References: <826@hou2b.UUCP>
Lines: 35
Nf-ID: #R:hou2b.UUCP:826:ccvaxa:5100132:000:1652
Nf-From: ccvaxa.UUCP!aglew    Sep 13 17:49:00 1986


... > Henry Spencer describes virtual memory as a technique for making
... > the hard limit of physical memory into a soft limit. 

Right. Consider these hypothetical graphs (I hope these come out
- I'm not using tabs):

Strictly physical memory system:
                             The cost of doing overlays
                        _/     (Chunked according to interconnectivity
                      _/       of program and data, and accuracy of
                     /         predictions)
       _____________/        Physical Memory Speed
	Size of Program/Data--->
    The vertical axis might be program speed or SW development cost.
                                                |
Virtual Memory System              _____________|
                     _____________/
     _______________/
           (penalty for virtual memory)
     ___________________________________________
		    ^                           ^
		    | Physical memory limit     | Virtual memory limit

For SW development, the penalty for virtual memory is nearly zero
(certainly zero for applications, but nonzero for OS, since somebody has
to write the memory management software). For execution, however, the
virtual memory penalty is appreciable - it may not slow you down, but it
will certainly take hardware and floorspace that you might have devoted
to something else. The question is, where do most of your applications lie
- below the physical memory boundary, or above? - and which ones cause people
to buy your system?

Andy "Krazy" Glew. Gould CSD-Urbana.    USEnet:  ihnp4!uiucdcs!ccvaxa!aglew
1101 E. University, Urbana, IL 61801    ARPAnet: aglew@gswd-vms