Ethernet and IP networking have gained nearly universal acceptance in offices, homes and wide-area connections,...
but they have not had the same success on the factory floor.
Some factory and manufacturing machines, such as computer-numerical-control equipment, demand guaranteed latency. Packets must arrive on a strict schedule. A delay of a fraction of a millisecond is unacceptable. As a result, the enterprise and factory floor have not been able to link via the same switched Ethernet or IP network.
Ethernet uses a contention-based access method, with no guarantee of how long a station can be delayed until it gains access to the network. Similarly, IP packets are sent with no delivery guarantee. The TCP layer ensures messages eventually arrive, but with no guarantee of how long it will take or that packets will arrive in order.
Virtual LAN (VLAN) priorities and quality-of-service guarantees ensure adequate performance for most enterprise network applications, but they cannot meet the requirements of the factory floor. Other applications have similar requirements. Now, a series of recent IEEE standards provides a way to connect the enterprise and the industrial control network, enabling both types of traffic to flow over the same network.
What is time-sensitive networking, and how does it work?
The IEEE work began to support audio and video studios. They have used special-purpose analog networks to provide the constant delay required by extremely high-quality sound and video. Each connection requires an individual cable, resulting in a confusing tangle. Replacing the tangle with a single Ethernet cable was an attractive option.
While the work began with a focus on audio and video needs, it soon became clear the factory floor and other applications shared the same requirements. The effort was broadened to develop a set of protocols that would apply more widely.
In 2012, the IEEE 802.1 Audio Video Bridging Task Group changed its name to the Time-Sensitive Networking Task Group. The IEEE task group developed a series of standards that enable time-sensitive traffic to meet strict latency requirements, while still allowing enterprise traffic to share the same network.
The standards define how time is split between time-sensitive traffic and nontime-sensitive enterprise traffic, with VLAN priorities used to differentiate traffic types. Each end-to-end packet flow is assigned to any of eight different VLAN priorities. All time-sensitive traffic is typically assigned to the highest-priority level.
Time-sensitive networking standards
IEEE 802.1Qbv specifies how time can be divided into fixed-length repeating cycles, with each VLAN priority level assigned to a time slot within each cycle. The VLAN containing time-sensitive traffic will be assigned a slot, while lower-priority, nontime-sensitive traffic is assigned to other slots. Each VLAN priority can be allocated to an individual time slot, or multiple priorities can be grouped in the same slot.
The length of each cycle must be configured to meet application latency requirements. Namely, the time slot allocated to time-sensitive data's VLAN priority must reoccur often enough to meet application requirements.
Strict traffic separation requires all switches and end nodes carrying time-sensitive traffic contain synchronized clocks. Packets cannot be delivered exactly when needed unless all elements in the network agree on time. IEEE 802.1AS specifies how synchronization is set up and maintained.
Time-sensitive packet flows must be carefully configured. The number of buffers in each node must be configured, as packet loss in a conventional network typically occurs because a switch or receiving node runs out of buffers. Packets cannot be sent too rapidly for the configured number of buffers. IEEE-defined packet-shaping algorithms determine the intervals between successive outgoing packets.
Maximum packet length is also configured, and the network path for each end-to-end flow is determined in advance. The result is no time-sensitive switch or end node will run out of buffers, packets will not back up in a switch queue, and they will not arrive out of order due to traveling different routes.
A guard band at the end of a lower-priority time slot prevents a lower-priority packet from interfering with the beginning of the slot allocated for higher-priority traffic. No transmission can begin during the guard band. A long packet that begins before the guard band -- and would extend beyond -- will be cut into two packet fragments, with the second sent after the time-sensitive time slot is over.
Revving up time-sensitive technology
The IEEE time-sensitive networking standards are now being applied beyond the initial applications. Electrical power generation and distribution currently rely on complex control networks. Replacing current technology with Ethernet offers a promising resolution.
The automobile industry is implementing time-sensitive technology to replace the bundle of individual cables in a vehicle. For example, functions such as engine timing and safety functions such as lane-departure warnings require tight limits on delay, but infotainment and navigation traffic do not. Combining these flows over the same Ethernet cable reduces the weight of multiple cables and the labor required to install them.
Work is now underway to extend time-sensitive networking standards to wireless networks. As current applications prove successful, additional applications will also benefit from this technology.