F3ARIoT: A framework for autonomic resilience of IoT applications in the Fog

Abstract

Fog Computing is especially appealing to the Internet of Things (IoT) because it provides computing, storage, and communication resources at the edge of the network, near the physical world (PW). Thus, IoT located in the PW can have interesting properties such as low latencies, real-time operations, and data privacy. The Fog, however, is unstable because it is constituted of billions of devices in a dynamic environment. Moreover, the Fog is cyber-physical and devices are thus subjected to external PW conditions which increase the occurrence of failures. When failures occur in such an environment, the resulting consequences on the PW can be hazardous and costly. This paper presents F 3 ARIoT, a framework for autonomic resilience of IoT applications in the Fog. This framework recovers from failures as well as maintains consistency and safety with respect to the PW during the recovery procedure. F 3 ARIoT was implemented and evaluated on a smart home application. A performance evaluation showed that it has a negligible overhead and recovers from failures in a very short delay with respect to end-users.

Hybrid Partitioning for Embedded and Distributed CNNs Inference on Edge Devices

“…Among them, energy cutoffs can occur, as is the case with customers who abruptly unplug their internet gateway to save energy. HyPS offers resilience to NN inference execution at the edge by enabling backups policies, such as those described and used for the IoT in [18].…”