Fog based energy efficient ubiquitous systems

Abstract

Growing ubiquitous computing systems generate large data that require large processing power. The data must be processed such that result is generated within a predefined time and also minimize the energy consumption of the system. The Cloud has sufficient processing power to process the data, but forwarding this data to the Cloud consumes huge energy. Fog environment solves this problem by performing processing near the ubiquitous computing system. Machine learning techniques used in the ubiquitous computing applications are divided into steps which can be performed at different layers. Our work considers a layered architecture comprising of Edge, Fog, and Cloud layers. The power consumed in processing and forwarding the data by devices at different layers is different. The challenge now arises to identify which machine learning steps to be performed at which layer such that energy consumption is minimized and the delay encountered remains within a given threshold. Mathematical analysis is presented for energy consumption by different devices at different layers. © 2018 IEEE.