Analysis of Energy Consumption in a Federated Learning-Based Zero-Touch Network

The current world revolves around data. Internet predicts that there are currently 2.8 million devices connected to the Internet. It spans the largest web with almost six connected devices per person. Cloud infrastructure is reaching its maximum capacity and hence needs upgrading. Fog Computing is a viable addition. This infrastructure upgrade also includes a suitable routing algorithm and its complement switching topologies. Adding self-learning capabilities to such a network implies the notion of Zero-Touch Networks. A pivotal point in Zero-Touch Networks is the selection of an optimal machine-learning algorithm. One such algorithm is Federated learning which relies on local updates of a global model. This paper revolves around, the analysis of energy consumption of a Federated learning-based utility supervision architecture for Zero-Touch Networks, by comparing it to a Cloud-Fog architecture. The initial test results were obtained through simulation on ‘iFogSim’. The analysis utilizes linear regression for prediction. The results depict lower initial values and less variation in a Federated learning-based architecture, as compared to Cloud-Fog architecture.