Real-Time Publish/Subscribe for the Industrial Internet of Things

点击翻译按钮获取中文摘要

The project RTiPS-4-IIoT deals with the Industrial Internet of Things (IIoT) in a future highly-flexible smart factory in that all information technology is seamlessly interconnected and employees, robots, and artificial intelligence cooperate tightly. Potential applications in a smart factory range from control of production having hard real-time constraints, over augmented reality environments with soft real-time constraints to best effort traffic, which connects operational information systems and office information technology. Supporting this mix of real-time constraints with varying strictness in a converged network infrastructure is challenging.For this scenario, a converged network infrastructure should offer flexible and decoupled m:n communication, while supporting several classes of real-time traffic at the same time. For an implementation, publish/subscribe communication is a good choice, as it is promoted and already used by standards, products, and projects in the area of the IIoT. However, existing approaches do currently not support hard real-time, because the flexibility of publish/subscribe is difficult to combine with the predictability necessary for hard real-time constraints. But this situation has changed with the advent of software defined networking (SDN) and time sensitive networking (TSN). With these new technologies, this combination now becomes practicable and also very promising. The project, therefore, researches a convergent network infrastructure that supports application-specific real-time constraints with varying strictness and that is implemented using publish/subscribe and software-defined networking. First, modeling real-time constraints in the context of publish/subscribe communication is researched. The developed models are, then, the basis for planning the communication relationshipsand for implementing them into the network using SDN and TSN. Later in the project, approaches to reconfigure a running system without sacrifying already guaranteed real-time constraints are investigated and it is ensured that planning also scales to larger scenarios.