TSN: The “Magical” Transformation of Ethernet

These days, so much of our work depends on data. It not only delivers instructions and guidelines for human actions, but often plays a direct role in decision-making and in issuing commands to machines. This is especially so in areas that are seen as relatively "critical", such as industrial manufacturing.

There is a saying, “if you want to get rich, build roads first”. Improving transportation infrastructure in the real world benefits production by facilitating the movement of people and logistics, thereby driving growth for the entire economic system. The same principle applies in the digital world. Only by providing an efficient network for the circulation of data can we unleash its tremendous power. However, while networks have existed for industry for some time, they have not yet achieved a high level of efficiency.

"Building roads" for industrial networks

The idea of building intersecting "roads" for data in the industrial field has been around for a long time. In the 1970s, with the development of programmable logic controllers (PLC) and automation technology, it gradually became necessary to implement distributed control of production equipment. This led to the development of industrial fieldbuses, which were capable of connecting previously isolated equipment. 

Every technology inevitably goes through a "Warring States" period during which differing standards compete for dominance in the early stages of development. The industrial fieldbus was no exception. Due to the unique "vertical" environmental structure in the industrial field, these standards have still not been unified to this day. This effectively limits the circulation of data, which obstructs the expansion of industrial networks and data sharing.

The challenges associated with the lack of integration among these differing standards escalated as the scale of industrial networks expanded. Therefore, in the early 21st century, "industrial Ethernet" was introduced. As the term implies, industrial Ethernet introduced Ethernet architecture, which was the worldwide standard in the IT field, to industry. The standard Ethernet medium was universally adopted, providing an opportunity to unify the standards on both the physical layer and data link layer.

In theory, this should have been an ideal solution. However, industrial Ethernet carries the "industrial" prefix because it differs from traditional Ethernet technology. Standard Ethernet uses the carrier-sense multiple access with collision detection (CSMA/CD) mechanism. When two data senders collide, they must wait for a predetermined amount of time before they retransmit their messages. This means that when there is network congestion, some packets may not be sent out for a long time. Essentially, Ethernet is a time-insensitive and uncertain network. For industrial automation control, timing is critical. To meet these requirements in industrial applications, standard Ethernet was “reformed” by the modification or addition of certain protocols to ensure absolutely reliable and precisely timed delivery of data, thus making it a deterministic network. This was the beginning of industrial Ethernet.

However, different manufacturers had their own ideas and methods for reforming the Ethernet. This resulted in a network that, while it was referred to as industrial Ethernet, was unable to achieve true interconnection and interoperability. The circulation of data through these sub-networks required additional gateways to be built. Imagine everyone coming together to build the same network of roads, but with different traffic rules existing on different roads and even on different sections of roads. Traffic could never flow smoothly on these roads – and nor could data flow smoothly through the sub-networks of industrial Ethernet.

An even greater challenge was presented by the emergence of smart manufacturing concepts such as Industry 4.0, which required a network that could support different types of data circulation at the same time. The data would have to be aggregated and processed in a unified network, regardless of whether the data is real-time data required for manufacturing site control or non-real-time data required for production management and optimization. Some of the overall optimization work could involve directly connecting to the edge and the cloud without passing through the traditional hierarchical controller. In other words, the ideal Ethernet architecture in industrial applications should be able to meet the operational technology (OT) network's real-time control requirements and support the convergent network required for high-throughput data transmission in the information technology (IT) field. This would enable the data of the entire industrial production process to be aggregated, eliminating all information blind spots.

Clearly, the realization of this ideal required in-depth reforms on traditional Ethernet to create a unified network and protocol specification to meet the requirements of the manufacturing industry. This was the only way to remove the barriers that prevented interoperability between different bus standards. And so Time Sensitive Networks (TSN) were born.

The Road to TSN

Although the application of TSN in the industrial field is much talked-about, it actually originated from a breakthrough in the audio and video field. The IEEE 802.1 working group established the Audio and Video Bridging (AVB) Task Group in 2006 to find effective solutions to the problems associated with the real-time synchronized transmission of data in audio and video networks. It soon became apparent that the challenges being addressed by the AVB Task Group had a lot in common with the time and deterministic problems of Ethernet data transmission. Thus, in 2012, the AVB Task Group expanded its scope of application and changed its name to the "TSN Task Group".

There were three main challenges that TSN had to address: time synchronization, scheduling and traffic shaping, and the selection, reservation, and fault tolerance of communication paths.

• Time synchronization: The implementation of time synchronization provides a common time reference for all devices in the network, thereby allowing clocks to be synchronized with each other. It also provides a negotiated benchmark for end-to-end transmission delays.
   
• Scheduling and traffic shaping: This allows the coexistence of different priority traffic categories on the same network, where each category has different requirements for available bandwidth and end-to-end delays, thus enabling the transmission of different data types on the same network.
   
• Selection, reservation, and fault tolerance of communication paths: This defines how all devices participating in real-time communication need to follow the same rules in selecting communication paths and reserving bandwidth and time slots, so that multiple paths can be used for troubleshooting. This ensures network security and reliability.

After years of development, TSN has defined a series of sub-standards that fall under the IEEE802.1 standard. This has resulted in a comprehensive set of protocols, which have provided a general processing mechanism for the MAC layer of the Ethernet protocol, and ensured the time certainty (real-time) of Ethernet data communication. It also enables interoperability between different network protocols.
,