Analytical models for availability evaluation of edge and fog computing nodes

Abstract

Although cloud computing environments increase availability, reliability, and performance, many emerging technologies demand latency-aware networks for real-time data processing. For instance, the Internet of Things environments are composed of many connected devices that generate data for applications, where many of them are latency-sensitive, such as facial recognition security systems in airports or train stations. To overcome the latency of the cloud infrastructure, researchers introduced the edge and fog computing paradigms in order to increase computing power between the cloud and devices. In this study, we propose analytical availability models; also, we evaluate the availability of physical edge and fog nodes running applications. To finish, we perform a capacity-oriented availability and a cost evaluation comparing edge and fog environments. Some of the results show that we can improve the availability from 2.96 number of nines to 5.93, by using our analytical models to plan the infrastructure. These models aim at supporting engineers and analysts to plan fault-tolerant edge and fog environments.