Advanced Analytics as a Service in Smart Factories

A key aspect of Intelligent Manufacturing is the interface between the Edge and Fog layers on one side and the Cloud on the other side. Once that some data extraction and cleaning has been performed in the (logical and/or physical) nearby of the production line, it is necessary to move to the Cloud where distributed data and parallel computations make it possible to train and use Machine Learning models. For a wide range of applications, it is also necessary to restructure and reconfigure the computation network over the time, reacting to relevant events at real-time. This can be achieved (i) by making the network and all of its components fully configurable through API calls, (ii) by using containerization technologies, and (iii) by relying on graphical user interfaces for development, data visualization, monitoring, and interaction. In this work, an architecture for a computation network is presented that (i) spreads across the Edge, Fog and Cloud layers, (ii) is fully configurable through API calls, and (iii) is based on Docker, Node-RED, MQTT, Scala, Spark, and Cloud storage technologies such as Hadoop. Two proofs of concept are presented: a clustering-based alerter, and a demo environment plus a Postman collection including over 600 API calls to demonstrate how the proposed architecture enables Big Data stream transformations and analytics as a Service.