Three Aspects of Real-Time Multiprocessor Scheduling: Timeliness, Fault Tolerance, Mixed Criticality

The design of real-time systems faces two important challenges: incorporating more functions/services on existing hardware to make the system more attractive to the mar-ket, and deploying existing software on multiprocessors (e.g., multicore) to utilize more processing power. Adding more services on the same hardware needs efficient resource utilization. In addition, satisfying the real-time constraints, while at the same time effi-ciently utilizing the multiprocessor platform, is a challenging problem. This thesis deals with global multiprocessor scheduling for real-time systems, that is, the fixed-priority scheduling of sporadic tasks, where each task is allowed to run on any processor. More specifically, this thesis considers three aspects of the design and analysis of global scheduling algorithms: timeliness, fault tolerance, and mixed criticality. Timeli-ness is about meeting the deadlines of the tasks; fault tolerance is about producing the correct output within the deadline even in the presence of faults; and mixed criticality is about facilitating the certification of systems when tasks having different criticality (or importance) are hosted on a common computing platform