Energy-aware Real-time Task Partitioning on Multi-core Processors with Shared Resources

Abstract

Nowadays, multi-core processors are increasingly adopted in embedded systems. These processors can achieve energy consumption minimization by employing dynamic voltage/frequency scaling techniques (DVFS). Several energy-aware real-time task partitioning algorithms have been suggested for multicore processors. While many of these algorithms focus on independent real-time tasks, there has been relatively limited research dedicated to task synchronization. This paper focuses on optimizing energy consumption by assigning dependent real-time tasks to a multi-core processor. When multiple tasks on different cores access shared resources simultaneously, it can result in longer blocking times, consequently increasing the execution time of tasks. This situation can result in missing hard deadlines, potentially causing system failure. The Highest Task-Based Partitioning (HTBP) algorithm is structured to decrease overall energy consumption while ensuring deadlines are met. It allocates tasks with high similarity (accessing the same set of resources) to the same core, effectively minimizing the occurrence of remote blockings. In the evaluation of the HTBP algorithm, we compared it with similarity-based partitioning (SBP), worst-fit decreasing (WFD) and best-fit decreasing (BFD). Our results indicate that our proposed (HTBP) algorithm outperforms SBP, WFD, and BFD algorithms (bin-packing algorithms), minimizes the overall energy dissipation, and improves schedulability.