A knowledge-based approach for real-time systems debugging

Abstract

The authors consider a method of testing unpredictable sequences known as program execution monitoring. In this method, a snapshot of the system events and happenings are recorded in a noninterfering manner, i.e. without corrupting the critical timing requirements of the target system. The traces thus collected are analyzed to isolate the bug in an offline mode. The examination of voluminous traces is a laborious and tedious task that requires a high degree of expertise. It is shown how this expertise can be encapsulated in a knowledge-based system and the examination process automated to at least localize the fault and answer questions about its reasoning. Such an approach shortens the overall debugging time and contribute toward software productivity and reliability.<>