Dataset for Partitioned Critical Real-Time Systems on Multiprocessor Platforms

This dataset corresponds to a databank of task sets for critical real-time systems partitioned on multiprocessor architectures. For its construction, synthetic task sets were generated and their offline planning was performed, extracting the metrics corresponding to each generated plan. A total of 5.76 million plans were generated. The objective of the dataset is to represent the behavior and characteristics resulting from applying different combinations of scheduling algorithms to different sets of tasks. For each scenario, 4,000 task sets were generated. These sets were initially processed by a partitioning task assignment algorithm using two approaches: mixed integer linear programming (MILP) based optimization and genetic algorithms (GA). Next, the assignment of tasks to cores was performed using ten different algorithms: MILP, GA, Worst-Case Decreasing Utilization (WFDU), Worst-Case Increasing Utilization (WFIU), Best-Case Decreasing Utilization (BFDU), Best-Case Increasing Utilization (BFIU), First-Fit Decreasing Utilization (FFDU), First-Fit Increasing Utilization (FFIU), as well as versions of WFDU and FFDU that consider task partitioning (WFDUPart and FFDUPart). Finally, with the tasks already assigned to the processors, planning was performed using three algorithms: Context Change Reduction (CC), Earliest Deadline First (EDF), and a Combined Planner (PC). From the generated plans, metrics were extracted to evaluate the quality of the planning and to determine whether the set of tasks was schedulable or not.