Browse Prior Art Database

Monitoring System Behavior In a Complex Computational Environment

IP.com Disclosure Number: IPCOM000128905D
Original Publication Date: 1979-Dec-31
Included in the Prior Art Database: 2005-Sep-20

Publishing Venue

Software Patent Institute

Related People

Mitchell L Model: AUTHOR [+3]

Abstract

Complex programming environments such as the representation systems constructed in Artificial Intelligence research present new kinds of difficulties for their users. A major part of program development involves debugging, but in a complex environment, the traditional tools and techniques available for this task are inadequate. Not only do traditional tools address state and process elemeuts at too low a conceptual level, but an Artificial Intelligence system typically imposes its own data and control structures on top of those of its implementation language, thereby evading the reach of traditional program-level debugging tools. This work is directed at the development of appropriate monitoring tools for complex systems, in particular, the representation systems of Artificial Intelligence research.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 1% of the total text.

Page 1 of 106

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

©; Copyright 1979 by Xerox Corporation; used with permission

Monitoring System Behavior In a Complex Computational Environment

by Mitchell L Model

CSL 79-1 JANUARY 1979

Abstract: See next page

This report reproduces a dissertation submitted to the Department of Computer Science and the Committee on Graduate Studies of Stanford University in partial fulfillment of the requirements for the degree of Doctor of Philosophy. It is also available as Stanford University Computer Science Department Report CS-79-701.

CR Categories: 3.69, 4.42, 8.2.

Key words and phrases: Debugging, complex systems, Artificial Intelligence languages, graphics.

XEROX PALO ALTO RESEARCH CENTER 3333 Coyote Hili Road / Palo Alto / California © 1979 by Mitchell L Model

Abstract

Complex programming environments such as the representation systems constructed in Artificial Intelligence research present new kinds of difficulties for their users. A major part of program development involves debugging, but in a complex environment, the traditional tools and techniques available for this task are inadequate. Not only do traditional tools address state and process elemeuts at too low a conceptual level, but an Artificial Intelligence system typically imposes its own data and control structures on top of those of its implementation language, thereby evading the reach of traditional program-level debugging tools. This work is directed at the development of appropriate monitoring tools for complex systems, in particular, the representation systems of Artificial Intelligence research.

The first half of this worn provides the foundation for the design approach put forth and demonstrated in the second. Certain facts concerning limitations on human information processing abilities which formed the background for much of the research are introduced. The nature of computer programs is discussed, and a concept of "computational behavior" defined. A thematic survey of traditional debugging tools is presented, followed by a summary of recent work. Observation of program behavior ("monitoring") is shown to be the main function of most debugging tools and techniques. Concluding this first part is an analysis of the particular difficulties involved in monitoring the behavior of programs in large and complex AI systems.

The second half presents an approach to the design of monitoring facilities for complex systems. The need for system-level tools similar to the ones traditionally available is indicated. A new concept called "meta-monitoring" replaces traditional dumps and traces with selective reporting of high- level information about computations. The importance of the visually- oriented analogical presentation of high-level information and the need to take into account differences between states and active processes are stressed. A generalized method for generating descriptions of system activity is developed. This met...