Time-Sensitive Networking in automation
Where are we now?
By Volker E. Goller
Today, those working in the industrial communications industry are destined to confront Time-Sensitive Networking (TSN). TSN is coming, and while there are questions about when and in what form it will come, it is unquestionable that the technology will have a profound impact on industrial communications and automation. However, for many, the advantages of TSN are not clear.
Introduced to offices in the early 1980s, Ethernet quickly became popular due to its (at that time) high throughput of 10 Mbps. However, this iteration of Ethernet was not practical for real-time applications, because it used a party line, and collisions occurring at high utilization rates caused problems in office settings.
Collisions were eliminated in the next stage of development through the introduction of switched networks. Additionally, quality of service (QoS) brought Ethernet datagram prioritization. However, even with QoS, standard Ethernet can only guarantee latencies up to a certain point, especially with high network utilization-making it unsuitable for industrial applications, which rely on guaranteed latency.
For high-priority datagrams for industrial applications, there must be guaranteed available bandwidth and buffer space. Standard Ethernet cannot provide that, because of the store-and-forward strategy commonly used in commercial multiport switches and the impossibility of reserving bandwidth.
Store and forward means that a switch receives a complete datagram before forwarding it. This has advantages in terms of processing in the switch, but also brings potential problems that can negatively affect latency and reliability:
- When going through a switch, a datagram is delayed by an amount depending on its length. If switches are cascaded, the effect is magnified.
- Because a switch does not have an infinite storage capacity, it can reject datagrams if the network is experiencing overutilization. This means that datagrams-even those given higher priority-can simply be lost.
- Long datagrams can block a port for relatively long times.
From the beginning, switch cascading posed a challenge in industrial environments. Apart from the star topology used in the information technology (IT) field, line, ring, and tree topologies are frequently used in automation. These adapted topologies significantly reduce Ethernet installation wiring requirements and costs. Therefore, in industry, two-port switches employing a cut-through strategy, where datagrams are forwarded before being completely received, are integrated into field devices.
One size fits it all
Because standard Ethernet did not have sufficient bandwidth reservation capabilities, automation experts began developing their own Ethernet extensions in 2000. However, paths diverged during development. There is differentiation between the following approaches:
- Protocols using Ethernet as a transport medium for a fieldbus: These protocols claim complete control over the Ethernet medium for themselves. Classic TCP/IP communications are only possible in piggyback style via the fieldbus (EtherCAT and POWERLINK) or through a channel assigned by the fieldbus (Sercos). Bandwidth control is firmly in the hands of the fieldbus.
- Protocols that guarantee bandwidth reservation through a time-slicing procedure on the Ethernet: PROFINET IRT should be mentioned here. IRT enables hard deterministic, real-time data transmission on the same cable on which soft real-time or background traffic is operated. A precise timing model for the transmission paths is necessary for planning the time slices.
- Protocols based on sharing of the Ethernet cable: These protocols use QoS and are at home in factory and process automation applications. PROFINET RT and EtherNet/IP are noteworthy examples. These protocols are limited to the range of soft real time (cycle time greater than or equal to 1 ms).
For these standards, special hardware support and, thus, special ASICs are needed. Because PROFINET RT and EtherNet/IP are also based on the embedded two-port switch with cut through, they are not exempt here. Flexible, hardware-based multiprotocol solutions solve the problem in an elegant manner.
Breaking free of past limitations, TSN extensions for standard Ethernet in accordance with IEEE 802.1 have successfully been developed. Thus, there is now a standardized layer 2 in the ISO seven-layer model with upward compatibility to the previous Ethernet and hard real-time capability. With 802.1AS-rev, TSN also defines an interoperable, uniform method for synchronizing distributed clocks in the network. Because best-effort communication always takes place with TSN, the common use of a cable is possible for hard real-time applications, as well as all other applications (e.g., Web server, SSH). TSN is not dissimilar to PROFINET IRT in that regard, and it also has comparable performance.
New with TSN is the need for more extensive network configuration. Centralized or decentralized configuration is possible, and both are currently being discussed and implemented. Interoperability between the two configuration mechanisms is a future development goal.
Timing model: PHYs, cables, and switches contribute to delays in data transmission. This must be considered with the time-slot method (PROFINET IRT and TSN time aware shaper [TAS]).
Practical advantages of TSN?
TSN will be used in building automation and the automotive industry in the future. As a matter of fact, the market for embedded TSN solutions is expected to be significantly bigger than the current market for all industrial Ethernet solutions put together.
This is because the greatest technical advantage of TSN over previous industrial Ethernet methods is its scalability. Unlike current industrial networks, TSN is not defined for a specific transmission rate. TSN can be used for 100 Mbps just as for 1 Gbps, 10 Mbps, or 5 Gbps. It also optimizes topologies, because adapted data rates can be selected for various segments. Whether it is Gbps, 100 Mbps, or 10 Mbps, a unified layer 2-IEEE 802.1/TSN-is used.
A uniform network infrastructure also helps personnel tasked with setting up and maintaining the network, because TSN solutions can now be used in sectors other than automation: building, process, and factory automation and energy distribution alike.
Training with TSN is underway to prepare the technology workforce for its pending widespread adoption. TSN is already a topic at many universities, mostly in the research stage. However, technical and vocational colleges are already showing interest in this topic. TSN will become basic knowledge for engineers, technicians, and skilled workers. Retraining for different fieldbuses will no longer be necessary.
Brownfield, or what will happen to today's protocols?
There is a recurring theme in nearly all TSN-related working groups: How do we safeguard the transition to TSN and the supply to existing installations, such as brownfield applications? Emphasis is being placed on making the transition to TSN easy for customers. Existing industrial Ethernet protocols are not just going to vanish overnight. Companies using PROFINET, EtherNet/IP, EtherCAT, or a similarly widespread industrial Ethernet protocol today can safely assume they will also be able to operate networks with these protocols-and receive support and replacement parts-in 10 years' time.
All industrial Ethernet organizations have models that describe how existing plants can cooperate with new TSN-based devices. The interface to the existing industrial network is made by a gateway (Sercos), with a coupler (EtherCAT), or without any special hardware (PROFINET RT). Especially PROFINET and EtherNet/IP plan to make their complete protocols available right on TSN as layer 2. This makes stepwise transition to TSN possible.
TSN as an opportunity
TSN will be found in new installations everywhere, as well as in the form of islands or segments introduced incrementally into existing installations. However, with TSN, there will be new players in the industrial Ethernet field. OPC UA, with the new transport protocol PUB/SUB, in conjunction with TSN, is already viewed as a competitor to the classic protocols. For the manufacturers of field devices, this means that the classic industrial Ethernet solutions, as well as TSN and the new players will have to be supported.
TSN makes it possible to create a uniform basis for all industrial communications. Once TSN is introduced, layers 1, 2, and 3 of the ISO seven-layer model will be unified in industry. Completely new levels of scalability and performance will be possible.