Fault-Tolerant Scheduling Of Real-Time Tasks In Heterogeneous Systems

Heterogeneous processors are widely used in various real-time systems because of its high performance and low energy consumption,such as flight control systems.In real-time systems,if tasks cannot be completed before deadline,it will cause disastrous consequences.To avoid the occurrence of missed deadline due to system failures,it is important to provide fault-tolerance for systems.Fault-tolerant scheduling is an effective way to achieve system fault-tolerance.System failures can be divided into two types: transient failures and permanent failures.This thesis has conducted in-depth research on these two failure types.The main work is as follows:First,we study how to tolerate transient failures.The existing transient fault-tolerant algorithms can support fault-tolerance for systems,but neglect real-time requirement of tasks.Therefore,we embed fault-tolerance and time constraints into the scheduling problem,and propose a deadline-based transient fault-tolerant scheduling algorithm DB-FTSA.It can calculate the number of tasks which can be tolerated based on a given deadline.It gives priority to high-priority tasks when providing fault-tolerance.Experimental results show that the proposed algorithm can guarantee high scheduling rate and can effectively improve system reliability.Second,we study how to tolerate permanent failures.Existing permanent fault-tolerant algorithms blindly back up the tasks and neglect the task's time constraints.Therefore,we propose a deadline-based permanent fault-tolerant scheduling algorithm.The algorithm quantifies the number of permanent faults that the system can tolerate according to a given deadline.To a certain extent,it avoids the situation that blindly backing up for tasks makes tasks miss deadlines.Experimental results show that the proposed algorithm can tolerate system faults to the most extend under the condition of meeting the time constraint.