DES based Modeling and Fault Diagnosis in Safety-critical Semi-Partitioned Real-time Systems

Abstract

Real-time safety-critical applications are increasingly being implemented on multiprocessor platforms and are generally scheduled using either partitioned or global approaches. Recently, semi-partitioned approaches are gaining importance due to their ability to leverage the merits of both partitioned and global scheduling. Given a multiprocessor system for executing a set of safety-critical applications scheduled using a semi-partitioned scheme, it is necessary to ensure proper functioning of the system even in the presence of processor failures. In order to avoid possible catastrophic consequences, the presence of processor failures must be timely detected and identified to take any necessary corrective action. Formal model based safe design methodologies such as Discrete Event Systems (DESs) are often desirable in the construction of sound and complete fault diagnosis mechanisms. This paper attempts to develop a DES based novel modeling framework for a multiprocessor system scheduled using an optimal semi-partitioned scheduler and provides a formal DES based fault diagnosis scheme to diagnose the presence of processor failures in the system. The proposed framework has been illustrated using a 3-processor system and its diagnosability analysis has been carried-out.