A real-time validator

Testing real-time systems under probabilistically distributed stimuli helps uncover faults that elude other forms of testing. In such testing, it is highly desirable that failures be detected rapidly, so that relevant internal states and traces can be collected. This paper describes a real-time validator which monitors inputs and outputs to the system under test and determines in real-time that a failure has occurred. It presents the key component of the validator, the models of beliefs about the legal evolutions of system behavior. The model is applicable to real-time systems whose external behavior is specified through communicating finite state machine models. The paper presents the derivation of belief models from the specification model and describes a more efficient, rule-based representation of such models. At the end, the paper describes experience with one implementation of a belief-based validator and its use in failure data collection for a small telephone exchange.