Fault diagnosis in computing networks

Thesis/Dissertation · Wed Jan 01 04:00:00 EST 1986

OSTI ID:6935908

Kreutzer, S E

This dissertation is concerned with system-level fault diagnosis, which is the problem of identifying faulty units (components) in computer systems. For the purpose of generality, a computer system, regardless of whether it is a computer network, multiprocessor computer, or distributed system, is considered to be a collection of units. Each unit is tested by other units within the system and the results of the tests are collected in order to identify the faulty units. The units and test assignments are represented by a directed graph, which is called a testing graph. In Chapter 1, motivations for studying the Theory of system-level fault diagnosis is explained. Chapter 2 describes the various models that are used to represent test assignments and the implications of test results, and some of the relevant previous research is discussed. Chapter 3 addresses the problem of determining the diagnosability number (t) of a system's testing graph, which is the maximum number of faulty units that the system can tolerate and remain diagnosable. For previous results, it is assumed that there is a central controller that collects the test results and makes the diagnosis. In Chapters 4 and 5 the author considers distributed diagnosis applicable to systems that do not have a central controller. In distributed diagnosis, each unit in the system arrives at its own diagnosis.

• Research Organization: Northwestern Univ., Evanston, IL (USA)
• OSTI ID: 6935908
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
COMPUTER NETWORKS
RELIABILITY
TESTING