Fault tolerant real-time global scheduling on multiprocessors

Many real-time multiprocessor scheduling techniques have been proposed to guarantee the timely execution of periodic preemptive real-time tasks. However timeliness is usually only guaranteed in the absence of faults, which may be unacceptable for some critical systems. We therefore address the problem of multiprocessor scheduling for preemptive real-time tasks so that the timeliness of the system can be guaranteed even in the presence of faults. This work focuses on global scheduling where tasks can migrate across processors. We consider two varieties of global multiprocessor scheduling: in the frame-based model, an aperiodic task set is scheduled to create a template (frame), and that schedule may be executed periodically. In the periodic model, each task in the set has a separate period, and is executed with no explicitly predetermined schedule. For each model, we show how to guarantee timely execution and recovery in the general case. We also propose solutions that improve upon this general case when all tasks require the same amount of time to recover from a fault.