Evaluation of deterministic fault injection for fault-tolerant protocol testing

Abstract

A deterministic test strategy consisting of deterministic fault injection at the message level is investigated. Messages sent by faulty units are replaced by such wrong messages that cause all program parts of the faultless protocol units to be executed subsequently. Since this well-aimed fault injection poses complex problems, heuristics based on the program flow of previous injections of wrong messages is dynamically applied. The program parts to be tested are selected with increasing granularity until either a design error is found or sufficient structural coverage is reached, which reflects the portion of tested program parts. Using a simplified program model, an algebraic analysis of the structural coverage and the design error coverage, which is the probability to reveal an existing design error, is carried out. It is shown that fault-tolerant protocol testing by deterministic fault injection achieve better coverages than by random fault injection.<>