Design and implementation of an edge computing platform architecture using Docker and Kubernetes for machine learning

Abstract

Huge data sets and high resources consumption are the prominent features of machine learning services. At present, machine learning services are often deployed on large-scaled cloud servers. The cloud utilizes its rich resources to perform the model training and prediction tasks, but the performance of this method is often limited by the unstable network conditions. To combine the rich-resources advantage of the cloud server with the stable-network performance of the edge computing technology, this paper proposes a Cloud-training and Edge-predicting framework. By integrating the Docker container technology and Kubernetes container choreography technology, we build an edge computing platform, and deploy a machine learning model (Inception V3) on the platform. With this method, we implemented machine learning services on the edge side. In this paper, we have described the designing and building process of the edge computing platform and the deployment procedure of the machine learning model in detail, and we have taken an experiment to implement the service to prove the feasibility of our ideas.