Analysis and optimization of fault-tolerant task scheduling on multiprocessor embedded systems
CODES+ISSS '11 Proceedings of the seventh IEEE/ACM/IFIP international conference on Hardware/software codesign and system synthesis
Towards fault-tolerant embedded systems with imperfect fault detection
Proceedings of the 49th Annual Design Automation Conference
Synthesis of communication schedules for TTEthernet-based mixed-criticality systems
Proceedings of the eighth IEEE/ACM/IFIP international conference on Hardware/software codesign and system synthesis
Quality-aware media scheduling on MPSoC platforms
Proceedings of the Conference on Design, Automation and Test in Europe
Aging-aware hardware-software task partitioning for reliable reconfigurable multiprocessor systems
Proceedings of the 2013 International Conference on Compilers, Architectures and Synthesis for Embedded Systems
Hi-index | 0.00 |
In this paper we are interested in mixed hard/soft real-time fault-tolerant applications mapped on distributed heterogeneous architectures. We use the Earliest Deadline First (EDF) scheduling for the hard real-time tasks and the Constant Bandwidth Server (CBS) for the soft tasks. The bandwidth reserved for the servers determines the quality of service (QoS) for soft tasks. CBS enforces temporal isolation, such that soft task overruns do not affect the timing guarantees of hard tasks. Transient faults in hard tasks are tolerated using checkpointing with rollback recovery. We have proposed a Tabu Search-based approach for task mapping and CBS bandwidth reservation, such that the deadlines for the hard tasks are satisfied, even in the case of transient faults, and the QoS for the soft tasks is maximized. Researchers have used fixed execution time models, such as the worst-case execution times for hard tasks and average execution times for soft tasks. However, we show that by using stochastic execution times for soft tasks, significant improvements can be obtained. The proposed strategy has been evaluated using an extensive set of benchmarks.