TSN is a set of standards, each specifying specific system behavior. The fundamental mechanisms of TSN are time synchronization, traffic scheduling, and system configuration. There are more TSN enhancements on the road-map to address the market requirements, but not all features and protocols are finalized yet; however, current solutions are compatible with future evolution. Test plans are currently available for the fundamental mechanisms which are described below. TSN is addressing layer 2 capabilities of deterministic Ethernet networks. DetNet is addressing the layer 3 requirements and is currently in a standardization process.
A control system usually consist of multiple distributed controllers. For control loops, synchronization to a common time source of all involved network elements is fundamental. There are multiple IEEE protocols available that are relevant for time synchronization.
IEEE 1588 Precision Time Protocol (PTP)
PTP is a protocol where all nodes in a network synchronize their individual clocks to the one with the highest available quality. A Best MASTER Clock Algorithm (BMCA) is used to select the clock to which all nodes in the network have to synchronize to. Nodes advertise their clock capabilities in order to be ranked. If a new node joins later with higher clock accuracy, it will become the grand-master clock for all other nodes.
IEEE 802.1AS gPTP
IEEE 802.1AS is a protocol to optimize distributed time synchronization of standard Ethernet networks. One of the nodes in the networks acts as the grand-master clock which sends time synchronization messages that are used by the slaves to modify their local clock. A peer delay mechanism is used to measure path delay in the network.
IEEE 802.1AS-Rev is a revision of the 802.11AS-Rev: It addresses the requirements to seamless switch to a backup grand-master clock in case the link to the primary grand-master fails. This requires a redundant network path to the backup grand-master. It thus enables a higher degree of fault-tolerance for time-critical applications. In addition, support for concurrent multiple timescales is specified, which could be used by applications requiring an additional global clock timescale.
In a TSN network, time-sensitive traffic coexists with best-effort traffic. In order to meet the bandwidth and latency requirements of time-sensitive applications, there needs to be a mechanism for the identification and prioritization of data streams in network elements. The following TSN standards are addressing these capabilities:
In order to prioritize traffic that can meet bandwidth and latency requirements of applications, a protocol known as Forwarding and Queuing for Time-Sensitive Streams (FQTSS) was specified. It is based on the Credit Based Shaping (CBS) traffic shaping mechanism, whereby bandwidth will be reserved for streams of time-sensitive applications. This mechanism works well as long as enough bandwidth is available, but has its limitation regarding the prediction of traffic characteristics in congested networks.
Industrial control systems require guaranteed and predictable cycle times, so that processes can be synchronized. That is why IEEE specified an enhanced traffic scheduling protocol based on time-aware gates. The transmission gates are corresponding with traffic classes with which queues can be associated. A prerequisite for the time-aware traffic scheduling of stream is a PTP-based time synchronization.
All 802.1Q protocols are utilizing VLAN tags for the classification and identification of traffic classes.
In a TSN network, applications should be able to dynamically announce their data rate and latency requirements to the underlying network. Such stream requirements need to be centrally configured with the involved network elements, while different requirements from multiple streams can coexist.
Bridged networks can be configured using the Stream Reservation Protocol (SRP), allowing talkers to declare their network requirements including traffic class, latency boundaries, and data rates. Requirements are propagated via messages advertised by each hop in the network to the listener. Therein, each node has to be able to fulfill the required performance. The listener sends back a ready message to indicate the talker to start the transmission. This decentralized approach has some drawbacks that are addressed by protocols using centralized network configuration.