Policies for System-level Diagnosis in a Non-Hierarchic Distributed System

The paper presents a policy for system-level diagnosis and error confinement in distributed systems. The class of architectures to which the policy applies is defined, and the physical supports needed are described. It is proved that mechanisms presented permit reaching deterministic results of diagnosis in finite time (thus avoiding deadlock or starvation effects) and that system-level error confinement is effectively obtained. Mechanisms are implemented by means of software.