Fast Finding Optimal Redundancy to Satisfy Reliability Requirement for Safety-Critical Parallel Applications on Heterogeneous Distributed Automotive Systems
Lizan Wang, Jiang Zhu, Shujuan Tian, Tingrui Pei, Haolin Liu, Yinying Li · 2019
Automotive system is the safety-critical and cost-critical heterogeneous distributed embedded system. Task assignment based on replication fault-tolerance strategy is commonly used to satisfy the reliability requirement of directed acyclic graph (DAG)-based parallel automotive application. However, automotive system development is redundancy constrained due to the limited hardwares and resources. Finding optimal redundancy under the reliability requirement saves the hardware and resource costs, and ensures the driving safety for the automotive industry, but the process is time-consuming by using traditional exhaustive method. In this paper, we propose the fast task assignment for optimal redundancy (FTAOR) algorithm to find minimum redundancy of DAG-based parallel automotive application under the established reliability requirement. Unlike exhaustive method, the verification time is reduced by using the FTAOR algorithm, which predicts and eliminates most meaningless computations effectively. Experimental results on parallel automotive applications at different heterogeneity show that the proposed FTAOR algorithm generates lower redundancy compared with the state-of-the-art heuristic algorithms and increases the task size to 150 with a shorter computation time compared with exhaustive method.