Proactive vs reactive failure recovery assessment in combined fog-to-cloud (F2C) systems

Abstract

The increasing number of end user devices at the edge of the network, along with their ever increasing computing capacity, as well as the advances in Data Center technologies, paved the way for the generation of Internet of Things (IoT). Several IoT services have been deployed leveraging Cloud Computing and, more recently, Fog Computing. In order to enable efficient control of cloud and fog premises, Fog-to-Cloud (F2C) has been recently proposed as a distributed architecture for coordinated management of both fog and cloud resources. Certainly, many challenges remain unsolved in combined Fog-to-Cloud systems, mostly driven by the dynamicity and volatility imposed by edge devices, such as the recovery of failures at the edge of the network. Indeed, possible failures in computing commodities may be prohibitive for the achievement of the envisioned performance in F2C systems. In this work, we assess proactive and reactive strategies for failure recovery of network elements by modelling them as a Multidimensional Knapsack Problem (MKP) and study the impact of each one on several aspects such as service allocation time, recovery delay and computing resources load. The obtained results show the effect each strategy brings, thus concluding with some analysis on the recovery strategy best suiting distinct IoT scenarios.