Dependable Scheduling for Real-Time Workflows on Cyber–Physical Cloud Systems
Junlong Zhou,Jin Sun,Mingyue Zhang,Yue Ma
DOI: https://doi.org/10.1109/tii.2020.3011506
IF: 12.3
2021-11-01
IEEE Transactions on Industrial Informatics
Abstract:Cyber–physical cloud systems (CPCS) are integrations of cyber–physical systems (CPS) and cloud computing infrastructures. Integrating CPS into cloud computing infrastructures could improve the performance in many aspects. However, new reliability and security challenges are also introduced. This fact highlights the need to develop novel methodologies to tackle these challenges in CPCS. To this end, this article is oriented toward enhancing the soft-error reliability of real-time workflows on CPCS while satisfying the lifetime reliability, security, and real-time constraints. In this article, we propose a dependable algorithm for scheduling workflow applications on CPCS. The proposed algorithm uses slack to recover failed tasks and allows all tasks to share the available slack in the system. To improve soft-error reliability, the algorithm first determines the priority of tasks, then assigns the maximum frequency to each task, and finally assigns the recoveries to tasks dynamically. Slack also can be used to utilize security services for satisfying system security requirements. The lifetime reliability constraint is met by dynamically scaling down the operating frequency of low-priority tasks. Extensive experiments on real-world workflow benchmarks demonstrate that the proposed scheme reduces the probability of failure by up to $52.1%$ and improves the scheduling feasibility by up to $83.5%$ compared to a number of representative approaches.
automation & control systems,computer science, interdisciplinary applications,engineering, industrial