Cost-Effective Time-Redundancy Based Optimal Task Allocation for the Edge-Hub-Cloud Systems

Abstract

Shifting tasks from the cloud to the edge Internet of Things (IoT) devices facilitates real-time decision support, yielding rise to the emerging edge-hub-cloud computing architecture. However, as edge devices tend to be more vulnerable to errors due to environmental/technological factors, the reliability of the application can be compromised due to this shift. We propose an optimization and design exploration framework for task allocation which maximizes reliability while respecting latency and energy constraints in computation and communication tasks, as well as memory computation constraints. It considers varying tasks' vulnerability factors, per possible executing device, and allows for different task re-execution reliability approaches based on full or selective dual or triple re-execution of tasks. We validate and showcase the effectiveness of our approach through a real-life case study for power tower/line inspection using UAVs (Unmanned Aerial Vehicles). Evaluation results show that our framework is capable of optimizing the overall reliability of the targeted application while satisfying strict latency, energy, and memory constraints, allowing the designer to explore different reliability approaches in a short period of time.

Design of Fault-Tolerant Distributed Cyber–Physical Systems for Smart Environments

“…Indeed, nodes may fail for several reasons, spanning from battery discharges or lack of connectivity to damages caused by the surrounding environment (e.g., bad weather conditions or accidental human activity). However, even if there is a large body of literature on Fog/Edge computing [ [2], very few approaches have considered the planimetryaware CPS design [3], [4] and the hardening of CPSs [5], [6]. Indeed, we claim that these two requirements have to be jointly considered and in a cost-effective way by instantiating and configuring a distributed computing platform with the minimal number of the cheapest edge nodes.…”

Fine-Grained Vulnerability Analysis of Resource Constrained Neural Inference Accelerators

“…Existing redundancy techniques for soft error mitigation come with high cost on area, power, and performance, which edge devices cannot afford [6]. Consequently, it is important to explore the vulnerability of these devices at design time and apply selective redundancy at a fine-grained level, as needed [7]- [10]. Recent works that study the vulnerability in the presence of soft errors in various accelerators are reviewed in [4].…”