Task migration for fault-tolerance in mixed-criticality embedded systems

Prabhat Kumar Saraswat, Paul Pop, Jan Madsen
2009 ACM SIGBED Review  
In this paper we are interested in mixed-criticality embedded applications implemented on distributed architectures. Depending on their time-criticality, tasks can be hard or soft real-time and regarding safety-criticality, tasks can be fault-tolerant to transient faults, permanent faults, or have no dependability requirements. We use Earliest Deadline First (EDF) scheduling for the hard tasks and the Constant Bandwidth Server (CBS) for the soft tasks. The CBS parameters determine the quality
more » ... service (QoS) of soft tasks. Transient faults are tolerated using checkpointing with rollback recovery. For tolerating permanent faults in processors, we use task migration, i.e., restarting the safety-critical tasks on other processors. We propose a Greedy-based online heuristic for the migration of safety-critical tasks, in response to permanent faults, and the adjustment of CBS parameters on the target processors, such that the faults are tolerated, the deadlines for the hard real-time tasks are satisfied and the QoS for soft tasks is maximized. The proposed online adaptive approach has been evaluated using several synthetic benchmarks and a real-life case study.
doi:10.1145/1851340.1851348 fatcat:tbowusbk7zd3bdfko3rkyguwcu