You’re driving down the highway at top speed when an accident happens in front of you. Luckily your AI-powered autonomous driving system recognizes this and immediately kicks into action, executing emergency braking with all wheels slowing down in a well-controlled way. Delayed action with just one wheel would have been disastrous and inevitably caused the car to skid. That’s why time-sensitive packet networks are needed in modern cars to ensure the delivery of critical control data in a deterministic way.

Many other applications in industrial automation or aerospace demand packet networks with deterministic behavior for fast and reliable delivery of data with very low jitter. Time-sensitive infrastructure is also required to connect cellular networks or professional audio/video-processing devices. Unlike the example of the car above, those use cases connect over public networks, which are not prepared for time-sensitive traffic. How can communication service providers add this important, mission-critical capability? Let’s take a look at technologies for low-latency packet networks and then highlight how those technologies can be applied with a highly relevant use case.

From best-effort to prioritized and express traffic forwarding

The latency performance of frame switches and packet routers depends on the load situation. With high utilization, data is increasingly stored in queues and not immediately sent out, causing higher and indeterministic delay. 

Time-sensitive networking (TSN) has been developed for the delivery of critical traffic in a semi-deterministic way, applying technologies that reduce latency and minimize jitter. How does it ensure that the data passes a switch without suffering from congestion-created delay, even under a very high load situation?

There are various technologies to make data forwarding more predictable.

• Priorities can ensure important express traffic is forwarded before best-effort traffic. 
• Bandwidth limitations for low-priority traffic reduce the overall load. 
• Frame preemption discontinues the transmission of a large low-priority frame that has already started, sneaking in high-priority traffic. 
• Scheduling defines timeslots for different traffic classes, ensuring that traffic with lower priority can’t block high-priority traffic. 

The network reliability can further be improved by frame replication, path reservation and more advanced filtering options.

Delivering timing over time-sensitive networks

Distributed control processes require precise time information to synchronize. With IEEE 1588 Precision Time Protocol (PTP), a technology for the delivery of accurate time can be applied. A specific PTP profile has been developed for critical applications using TSN. It assures timing performance requirements, provides strategies to cope with master clock failures and supports multiple clock domains. 

How to apply TSN?

There are several ways to implement TSN as listed above. Depending on the specific use case, the most appropriate solution can be selected. If an application requires data at a predefined point in time, scheduling schemes might be most appropriate. If the amount of time-sensitive traffic is low, a preemption scheme might be the best choice. 
Let’s have a look at an innovative radio access network with radio units, distributed units and central units interconnected by a packet-optical transport network. The fronthaul interface Fx has stringent timing requirements, which can even limit the distance between the DU and RU site.

The below table is based on the eCPRI 2.0 specification. It provides maximum one-way frame delay for latency-sensitive applications. Those values range from very demanding 25µs for High25 services to 500µs for High500 services, while medium and low CoS can be satisfied with 1ms and 100ms respectively. Please also note that light travels at 0.2km/µs through fiber, which results in 20µs for 4km and allows for the packet devices to add 5µs delay if ultra-low-latency applications need to be supported.

Introducing TSN into our market-leading packet edge demarcation and aggregation portfolio

ADVA has been investigating TSN technologies for many years. In the EU-funded Horizon 2020 project 5G-PICTURE, we showcased a TSN network for smart city applications and verified it with very demanding massive-MIMO xRAN use cases. It was recognized with the best demo award as early as ECOC 2018.

We continued to improve this technology until it was recently released as a feature enhancement to our 100Gbit/s demarcation and aggregation portfolio. With the latest version of our FSP 150-XG418, we’ve released a 100Gbit/s demarcation and edge aggregation device for Ethernet and IP business services that supports TSN featuring MAC merge sublayer IEEE 802.3br capabilities. Tests in real-world use cases confirm the significant latency gains of traffic preemption, assuring link latency of 5µs in back-to-back configurations and 25µs over 4km of fiber. Hence, our implementation keeps the latency at ultra-low values even in highly congested operations and overloads with best-effort traffic.

Summary

Latency-sensitive applications demand TSN for fast, reliable and secure connectivity. Packet-forwarding can be made deterministic, minimizing the impact of even high loads of best-effort traffic on express traffic. Frame preemption is an efficient, standardized technology, which is now available with our market-leading 100Gbit/s demarcation and edge aggregation devices. Early demos have proven its effectiveness, and our enhanced products are now providing service providers and operators of enterprise networks with an efficient and seamless solution.