Fault-Tolerant Dynamic Scheduling of Object-Based Real-Time Tasks in Multiprocessor Systems

An adaptive scheme for fault-tolerant scheduling of soft real-time tasks in multiprocessor systems

“…Moreover, the issue of resource access control for shared resources complicates the synchronization issue further. These are the reasons why most of the fault-tolerant scheduling algorithms (e.g., [3][4][5]9,14]) for multiprocessor systems assume a non-preemptive scheduling model, as adopted in this paper.…”

“…Primary-backup (PB) scheduling is an important faulttolerant approach in which two versions of a task are scheduled on two different processors and an acceptance test is used to check the correctness of the result [3][4][5]9,11,14]. The three variants [5] of this approach are: primarybackup exclusive (PB-EXCL) [3,9], primary-backup concurrent (PB-CONCUR) [5,15], and primary-backup overlap (PB-OVER) [5].…”

“…The three variants [5] of this approach are: primarybackup exclusive (PB-EXCL) [3,9], primary-backup concurrent (PB-CONCUR) [5,15], and primary-backup overlap (PB-OVER) [5].…”

“…In most real-time applications, there is a need for reliable execution of tasks even in the presence of faults. Reliable (fault-tolerant) execution of a task is usually achieved by scheduling multiple versions of the task [1][2][3][4][5][6]9,11,14,15].…”

Controlling schedulability-reliability trade-offs in real-time systems

“…On the negative side, it requires the execution interval (i.e., the interval between start time of primary and finish time of backup) of a task to be at least twice that of the task execution time. This not only increases the chances of a task (backup) not being scheduled before its deadline, but also introduces holes in the schedule [26], thus resulting in poor schedulability. In PB-CONCUR approach, the primary and the backup versions of each task are executed concurrently.…”