Detection Of Errors Using Aspect-Oriented State Consistency Checks

Software systems have grown in size and complexity. It is not possible to exhaustively test them to a fault free condition before they are deployed. As a result, their execution may be marked by failures. Often, these failures are preceded by a deterioration of internal state of the program (errors) before an externally observable failure occurs. Operationally, it would be beneficial to provide an indication of the internal well-being or ‘health’ of the system. At minimum, the system operator could be provided with a warning, so that corrective action could be taken. An aspect-oriented approach for implementing health monitoring of software systems is presented in [1], along with a classification scheme for health indicators. This paper presents a health indicator which reflects the extent of corruption of the internal program state. The focus is on embedded real-time systems, whose top-level objects have finite state machine behavior.