The Design of Real-Time Fault Detectors

This paper presents the design and implementation of real-time fault detectors. We describe their design, implementation, and scheduling under a Fixed Priority/ High Priority First policy. Two types of real-time detectors are described; primary detectors and secondary (meta) detectors. A Primary Detector is designed for the detection of simple faults and failures (Worst Case Execution Time, Worst Case Response Time, Latest Response Time and Activation Overrun events). These events occur when a task uses more resources than have been catered for. The secondary type of detector, called

meta Detector

, is used to detect more complicated events called

meta-events

. Meta-events are based on a set of primary detectors and their interrelations. The Real-Time Specification Language (RTSL) is used for the description of Meta-events, including the primary events relations such as precedence; (THEN) and other logical relations; (AND, OR, TIMES). Primary and meta fault detectors must be admitted to the system as periodic or sporadic real-time threads. We present a method for the feasibility analysis of each detector type. These principles are integrated within a Minimum Real-Time CORBA prototype called RT-SORBET.