List scheduling algorithms for open distributed real-time embedded systems

In recent years, the field of embedded systems has evolved towards application areas that combine stringent real-time constraints, reliability requirements and a need for an open-world assumption. Such systems are called Open Distributed Real-time Embedded (ODRE) systems. These systems are based on an open-world assumption where new components enter at runtime to dynamically realise emerging services. At the same time, reliable operation and support for stringent real-time requirements are essential to support closed-loop control and guaranteed response times. In such systems, time-triggered scheduling of the services before executing them on the system ensures the correct temporal behaviour of the applications and guarantees support for stringent real-time constraints. However, scheduling a real-time application in an ODRE system is complicated because of its dynamic nature. The structure of the ODRE system is continuously changing, which means that a fixed schedule generated at the design time cannot be used throughout the lifetime of the system. The execution times of prevalent scheduling algorithms, e.g, evolutionary algorithms, mixed integer linear programming, satisfiablity modulo theories, are not suitable for invocation at runtime. Moreover, these algorithms make assumptions, e.g., assuming a bus-based communication network instead of multi-hop communication networks or considering the same period for all the application tasks, that are unrealistic for the stringent real-time requirements of ODRE systems. Therefore, there is a need for a scheduling algorithm that computes a feasible schedule for ODRE systems at runtime whenever there are changes in the system. This thesis proposes list scheduling algorithms that consider both stringent timing constraints and openness of the ODRE system while computing a feasible schedule with short scheduling delay at runtime. The proposed algorithms are generic and support the computation of a time-triggered schedule for any system application on an ODRE system. In addition, this thesis demonstrates the models and algorithms for scheduling diagnostic services for fault detection and diagnosis in ODRE systems. The scheduling algorithms are evaluated based on different parameters such as, the size of the system application, the number of available resources, the network topology used in the system, the number of modifications in the scheduled application, and reconfiguration cost of the system. The results show that the algorithms can compute a feasible schedule provided that there are enough resources in the system. The results for the proposed incremental scheduling algorithm show that the computed schedules are scalable to changes in the system and the revalidation efforts can be reduced by minimising the changes in the already scheduled application. Furthermore, the time complexity and the observed runtime of the algorithms shows that they can be invoked during runtime when the scheduler is triggered by a change in the system configuration.