Energy-Efficient Partial-Duplication Task Mapping Under Multiple DVFS Schemes

On multicore platforms, reliable task execution, as well low energy consumption, are essential. Dynamic Voltage/Frequency Scaling (DVFS) is typically used for savings, but with a negative impact on reliability, especially when the applied frequency low. Using high frequencies, required to meet reliability constraints, or replicating tasks increases consumption. To reduce while enhancing and satisfying real-time we propose hybrid approach that combines distinct enhancement techniques, under task-level, processor-level system-level DVFS. Our mapping problem jointly decides allocation, assignment, duplication, constraints. This achieved by formulating Mixed Integer Non-Linear Programming problem, equivalently transforming it into Linear Programming, can be optimally solved. From obtained results, proposed achieves better finding solutions, replication approaches fail.