Energy-efficient primary/backup scheduling techniques for heterogeneous multicore systems

In this paper, we consider energy-efficient and fault-tolerant scheduling of real-time tasks on heterogeneous multicore systems. Each task consists of a main copy and a backup copy which are scheduled on different cores, for fault tolerance purposes. Our framework deliberately delays the backup tasks in order to cancel them dynamically when the main task copies complete successfully (without faults). We identify and address two dimensions of the problem, i.e., partitioning tasks and determining processor voltage/frequency levels to minimize energy consumption. Our experimental results show that our proposed algorithms' performance levels are close to that of an ideal solution with optimal (but computationally prohibitive) partitioning and frequency assignment components.