Theory of Fault Tolerance.
Abstract
The first phase of a long range program to develop a theoretical base for the design and evaluation of fault tolerant digital systems is summarized. The purpose of this effort was to investigate the feasibility of representing fault tolerant phenomena with two existing Labeled Graph models, LOGOS and Petri Nets. LOGOS and Petri Net models were used to represent systems at a functional level. Functional faults were hypothesized to describe the effect of a fault mechanism from a functional viewpoint. Functional faults were found to provide high leverage over traditional 'Stuck-At-One' or 'Stuck-At-Zero' analysis. Models were developed which describe: (1) The non-faulty system function; (2) The faulty system operation; (3) Fault detection techniques; and (4) Fault recovery schemes. The results of modeling several candidate fault tolerant system problems reaffirmed our belief in the utility of Labeled Graphs as a representation language for fault tolerant phenomena. We found both models concisely described the structure and dynamics of a control system and provided insight into the effect of fault detection and recovery mechanisms on a given fault mechanism.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 22, 1975
- Accession Number
- ADA014935
Entities
People
- L. A. Jack
- M. D. Johnson
- W. L. Heimerdinger
Organizations
- Honeywell International, Inc.