Software architecture for distributed real-time embedded systems

Abstract

Real time embedded systems have particularly strict requirements on accuracy, safety and reliability. A central question in the design of such systems is how to support concurrent processing without adversely affecting the timing requirements of the system. Concurrent processing is essential because the only way to successfully meet some tight real time constraints is to use multiple processors. This thesis focuses on the distributed scheduling problem. It proposes a distributed scheduling algorithm to allocate and schedule a set of tasks onto a collection of processors linked by a network. It further proposes a distributed software architecture for CAPS (Computer Aided Prototyping System) generated prototypes based on GLADE (GNAT Library for Ada Distributed Execution) . The new distributed CAPS architecture is applied to several prototype examples. The results show that it is possible to build distributed real time embedded systems under the distributed scheduling model, where sets of tasks run independently on each processor, using GLADE.