Communications in Computer and Information Science

Availability of computation capabilities and real-time machine data is one key requirement of smart manufacturing systems. Latency, privacy and security issues of cloud computing for Industrial artificial intelligence (AI) led to the edge computing paradigm, where computation is performed close to the data source. As on-premise edge deployments require companies to allocate budget and human resources to acquire and maintain the required information technologies (IT) infrastructure and equipment, they are not feasible for several companies. However, 5G can merge advantages of previous alternatives. Multi-Access Edge Computing (MEC) servers deployed at the edge of the 5G network close to the final user, offer security, privacy, scalability, high throughput and low latency advantages. MECs are suitable for industrial AI, while industrial companies do not face the burden of acquiring and maintaining servers and communication infrastructures. This paper proposes a real-time high-frequency data streaming architecture to deploy Industrial AI applications at MECs. The architecture has been successfully validated with data sent through a 5G network to a Kafka broker at the MEC, where different microservices are deployed in a Kubernetes cluster. The performance of the architecture has been investigated to analyze the capabilities of 5G and MEC to cope with the requirements of Industrial AI applications.