Network Slicing in Industry 4.0 Applications: Abstraction Methods and End-to-End Analysis

Industry 4.0 introduces modern communication and computation technologies such as cloud computing and Internet of Things to industrial manufacturing systems. As a result, many devices, machines, and applications will rely on connectivity, while having different requirements to the network, ranging from high reliability and low latency to high data rates. Furthermore, these industrial networks will be highly heterogeneous, as they will feature a number of diverse communication technologies. Current technologies are not well suited for this scenario, which requires that the network is managed at an abstraction level, which is decoupled from the underlying technologies. In this paper, we consider network slicing as a mechanism to handle these challenges. We present methods for slicing deterministic and packet-switched industrial communication protocols, which simplify the manageability of heterogeneous networks with various application requirements. Furthermore, we show how to use network calculus to assess the end-to-end properties of the network slices.

I. Introduction

The fourth industrial revolution, known as Industry 4.0, brings cyber-physical systems and Internet of Things (IoT) to industrial manufacturing systems [1], [2]. Furthermore, the number of interconnected physical devices will increase drastically, and they will continuously interact with local cloud services in order to act intelligently and flexibly. This introduces numerous challenges to industrial networks, which have traditionally been very static and strongly isolated  [3], [4]. First, it is expected that many new technologies, comprising both wired and wireless connections, will gradually be introduced into production lines resulting in a very heterogeneous network [5]. Second, the network will have to serve a wide range of applications with different quality-of-service (QoS) requirements, ranging from traditional closed-loop control systems to event-driven sensors and augmented reality (AR) displays. For instance, control and alarm systems may require a delivery reliability in the order of and end-to-end latencies in the range of 0.5–5 ms, while, at the same time, interactive applications require high data rates and moderate latencies [6], [7]. Finally, the increased system complexity also poses a challenge in managing the network and, in particular, the end-to-end QoS. This necessitates programmability of the network, as well as a framework for analyzing the end-to-end network characteristics [3], [8].