Energy Efficient, Real-time and Reliable Task Deployment on NoC-based Multicores with DVFS

Abstract

Task deployment plays an important role in the overall system performance, especially for complex architectures, including several cores with Dynamic Voltage and Frequency Scaling (DVFS) and Network-on-Chips (NoC). Task deployment affects not only the energy consumption but also the real-time response and reliability of the system. In this work, a task deployment approach is proposed to optimize the overall system energy consumption, including computation of the cores and communication of the NoC, under task reliability and real-time constraints. More precisely, the task deployment approach combines task allocation and scheduling, frequency assignment, task duplication, and multipath data routing. The task deployment problem is formulated using mixed-integer non-linear programming. To find the optimal solution, the original problem is equivalently transformed to mixed-integer linear programming, and solved by state-of-theart solvers. Furthermore, a decomposition-based heuristic, with low computational complexity, is proposed to deal with scalability. Finally, extended simulations evaluate the proposed methods.

Contention and Reliability-Aware Energy Efficiency Task Mapping on NoC-Based MPSoCs

“…To improve system efficiency, such as achieving energy and communication efficiency and reliability, task mapping on NoCbased MPSoCs has gained significant attention in recent years. Optimization objectives RA CA LB [27] Energy × × × [16] Energy × × × [21] Energy √ × × [28] Task Replication Overhead √ × × [20] Energy √ × × [29] Energy + Temperature √ × × [6] Latency × √ × [30] Latency × √ × [31] Makespan × √ × [13] Makespan × √ × [32] Energy + Latency × √ × [15] Energy + Latency × √ × [14] Energy × √ × [18] Energy + Reliability + Latency √ × × [19] Energy + Reliability + Latency √ √ × [22] Communication Efficiency + Workload × × √ [33] Energy Energy efficiency is essential in NoC-based MPSoCs, especially for battery-powered mobile embedded systems [16]. Several works have studied energy-aware task scheduling on NoC-based MPSoCs [16], [27] to minimize the total energy of computation and communication under real-time constraints.…”

“…(4) We perform simulations on randomly generated task sets and realistic application task sets to evaluate the performance of our method in terms of energy consumption, task reliability, and communication contention. Compared to the state-of-the-art methods [6] and [21], our method achieves 31.6% and 21.7% energy reduction and has 95.5% and 98.6% lower average contention, while balancing workload and satisfying reliability constraints. In addition, compared to the optimal solution, our method has a polynomial computation time, but with only 24.8% lower energy efficiency.…”

“…Therefore, the trade-off between energy consumption and system reliability should be considered during task mapping on NoCbased MPSoCs. Some existing works [18]- [21] have studied the joint design problem of saving energy and enhancing system reliability. For instance, transient faults are considered on links and routers [18], [19] or cores [20], [21].…”

“…Some existing works [18]- [21] have studied the joint design problem of saving energy and enhancing system reliability. For instance, transient faults are considered on links and routers [18], [19] or cores [20], [21]. As transient faults may occur during task execution and communication, it is necessary to consider that they can occur on cores, routers, and links when optimizing task mapping for energy consumption and system reliability on NoC-based platforms.…”

“…For the NoC-based platforms, it is necessary to consider communication contention during task mapping. However, most existing works focused on energy and (or) reliability improvement but ignored the influence of communication contention [18], [21]. Other works alleviated the network congestion by introducing fixed slack time into the task scheduling [13].…”