Diagnosis, Self-Diagnosis and Roving Diagnosis in Distributed Digital Systems.
Abstract
This report studies various problems related to the diagnosis of systems which may be conveniently partitioned into subsystems on modules capable of performing tests among themselves. Traditionally, system diagnosis has aimed at the detection of one or more faulty units in systems. In certain critical operations like large distributed systems, it is desirable to know precisely which units in the system are fault-free. The report first considers the problem of determining at least one fault-free unit in a system. Necessary and sufficient conditions, as well as more convenient sufficient conditions for solving the problem have been obtained. The 'roving diagnosis' scheme introduced in the report is an attempt to achieve self-diagnosis in distributed digital systems while involving as few units in diagnosis as possible at any given time. A 'window' is associated with the subsystem composed of the testing and tested units. Diagnosis of the entire system is achieved by moving this window around the system. Related issues which are studied include reconfiguration of the system under a fault and the reusability of the system after reconfiguration. Certain properties of the system communications graph are shown to lead to desirable systems where the delay in reconfiguring around a fault is small.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1978
- Accession Number
- ADA069770
Entities
People
- Ravindra Kumar Nair
Organizations
- University of Illinois Urbana–Champaign