Dew Computing in Industrial Automation: Applying Machine Learning for Process Control

Network infrastructure in the era of Big data, Industry 4.0, IoT, and AI, pushed the boundaries of data availability, scalability, reliability, high bandwidth, and low latency. This resulted in a new paradigm of cloud computing providing centralized data storage, efficient resource management, and software products marketed in an as-a-service manner. However, there are still many QoS-sensitive use-cases where data and service availability, network latency, and throughput needs to be further addressed. Partial solutions to these problems came in Fog and Edge computing, providing computation and data closer to end-user network devices. But these concepts still do not fully address the issues of offline data availability and network latency. A potential answer to these questions could be the concept of dew computing - an additional layer in the existing client-server architecture, operating on end devices to achieve the highest possible level of data synchronization between data in the cloud and local devices, which in the standard Cloud-Fog-Edge architecture relies strictly on connectivity. This paper addresses the implementation of dew computing in the field of industrial automation and data analysis performed in the cloud, with local execution of the resulting model algorithm by reducing network latency and providing offline data availability, providing the necessary data synchronization between the cloud and local databases.