Edge Computing and Analytics: An Extended Systematic Mapping Study

Connected sensors and devices, the Internet of Things, already today produce more data than data connectivity and cloud services can handle. This gives rise to various forms of distributed sensor data handling, from surveillance cameras with built-in feature detection to alarm functionality in temperature sensors. There is an increased interest in being able to use this sensor data for distributed intelligence. The term "Edge Computing" is often used to denote this distributed computing, performed close to the sensors providing the data. Edge Computing enables services typically provided by cloud services with less communication, lower latency, and independence of the internet infrastructure. In this paper, we present a comprehensive, unbiased overview of state-of-the-art research on edge computing and analytics. From the taxonomy of the 90 identified articles, most articles address task scheduling and operation partitioning while data management and engineering, image and facial recognition, power optimization, and anomaly detection are generally also covered. Simulation remains the most used approach for validation, and research results based on implementations of edge systems in real-life environments are still sparse.