Task replication based energy management using random-weighted privacy-preserving distributed algorithm for real-time embedded system

Abstract

Efficient energy management in real-time embedded systems is increasingly challenging due to the growing complexity of distributed tasks and the need for robust privacy preservation. Conventional task mapping and repartitioning techniques have focused on increasing the system reliability, efficiency, and lifespan, but typically incurred a high peak power generation because of Thermal Design Power (TDP) limitations which confines the scalability and applicability. To overcome these problems, the Task Replication-based Energy Management using Random-weighted Privacy-preserving Distributed Algorithm (TR-EM-R-RWPPDA-RTES) is proposed as a new scheme for real-time embedded systems. This architecture integrates Hotspot-Aware Task Mapping (HATM) to optimally load tasks across cores, Dynamic Heterogeneous Earliest Finish Time (DHEFT) scheduling to improve execution timing, and a Reliability-based Random-Weighted Privacy-Preserving Distributed Algorithm (R-RWPPDA) to optimize power consumption. Using these elements, the proposed approach reduces both system energy consumption and system trustworthiness. Comprehensive simulations based on the MiBench benchmark suite, as well as gem5 and McPAT simulators on ARM multicore processors (4, 8, and 16 cores), are also shown to validate the robustness of the proposed method. TR-EM-R-RWPPDA-RTES yields 23.73 %, 36.33 %, and37.84 % peak power consumption reduction with respect to the state-of-the-art solutions, thus providing a robust solution for energy-efficient, robust and reliable real-time embedded systems.