Real-Time Dynamic Scheduling Algorithms for the Savings of Power Consumption and Fault Tolerance in Multi-Processor Computing Environment

Saving energy consumption of modern processors has recently become popular due to the fact that high power consumption increases heat dissipation, which leads to decreased reliability of systems. This paper aims to generate a task schedule that can fully exploit the dynamic scheduling support (as well as the underline voltage/frequency scaling capability) of the target machines such that the real-time constraints can be met with as minimum as possible energy consumption. The algorithm STFBA (shorted task first based algorithm) proposed utilizes the policy of shortest-task-first and other efficient techniques, such as shared slack reclamation, to save energy in homogeneous systems for independent task set and task set with relationship constraints, respectively. STFBA comprises static part and dynamic part to lower time complexity of algorithm. Furthermore, a dynamic fault-tolerance algorithm based on STFBA is proposed, which is combined with the policy of shortest-task-first and checkpoint, to tolerate transient faults during the executions of tasks, while meeting the requirement of timing constraints and reducing energy consumption efficiently. The feasibility condition of checkpoint placement and lower bound of static processing speed is given to reduce the fault-tolerance scheduling costs. Compared with the efficient algorithms presented so far, simulation results indicate that proposed algorithms have much better scheduling performance trade-off in terms of makespan and energy savings when the slack generated by tasks is sufficient.