Iterative robust multiprocessor scheduling

General purpose platforms are characterized by unpredictable timing behavior. Real-time schedules of tasks on general purpose platforms need to be robust against variations in task execution times. We define robustness in terms of the expected number of tasks that miss deadlines. We present an iterative robust scheduler that produces robust multiprocessor schedules of directed acyclic graphs with a low expected number of tasks that miss their deadlines. We experimentally show that this robust scheduler produces significantly more robust schedules in comparison to a scheduler using nominal execution times on both real world and synthetic test cases.