Timely transport of data isn’t a new topic. Didn’t we have a similar discussion some (30) years ago when synchronous transport networks replaced plesiochronous digital hierarchy (PDH) technology? Some of you might remember the problem we were facing at that time: Interfaces worked at slightly different clock rates resulting in dropped frames. This was addressed with the synchronous network technology SDN/SONET, which transported data but also delivered very precise frequency information.
With the rise of packet networks and IP, the need for synchronous transport disappeared. Initial data traffic was mainly best-effort, with very low timing requirements. Ethernet technology as the transport layer for IP started to replace SDH/SONET.
Today, Ethernet and OTN, as the SDH-successor, are the predominantly applied interface technologies for transport networks, even penetrating radio access networks because of their ubiquity and low cost. But those technologies weren’t originally designed for the delivery of precise synchronization information.
Was this a good idea? Over the last years, we observed an ever-increasing need for transport of frequency, phase, and time information over data networks. Mobile networks, video application, and industrial control require very accurate timing. With SyncE and Precision Time Protocol (PTP), two complementary technologies were introduced in Ethernet transport networks for accurate delivery of frequency and time information.
But user data still suffers from delay and delay variations, which are not acceptable for time critical applications. Data transport networks need additional means to transport critical traffic in a time-sensitive way.
Delay and delay variations in (packet) networks
What options do we have to reduce delay and delay variation in packet transport networks? Let’s start with looking at vehicles and traffic jams to explain the challenges and options we have in packet networks. Here are some of the methods used on the road to reduce vehicle delay and mitigate unpredictability of arrival time.
Over-provisioning: Building wider roads with more lanes is a very effective but expensive way to avoid traffic jams and to assure that any trip reaches its destination without delay.
Priorities: At toll stations, there are referenced lanes for those with onboard units for automated billing. Fire trucks headed to a call obviously get priority at crossings. Such prioritized traffic doesn’t need to queue, but can proceed without delay.
Reservation: In some communities, there are lanes reserved for specific vehicles such as taxis, busses and security services. They move fast and don’t suffer from congestion. This, however, requires a fair amount of investment as such expensive additional lanes serve a relatively low number of vehicles.
Scheduling and policing: A transport authority might limit the number of vehicles using a street or entering a tunnel. If executed at an access ramp to an autobahn, this might result in delays for entering the autobahn but could avoid traffic jams on the core transport infrastructure.
Preemption: Now let’s assume that the unavoidable has happened and congestion is slowing down traffic. Ambulance and other services, however, will still be able to come through as their way gets cleared by other road users. Such express traffic is “cutting through” the already existing traffic slowing it further down.
Time-dependent reservation: On the German autobahn, lorries are not allowed at weekends. This helps to avoid congestion, especially during school vacations with many people driving to their holiday destinations.
Deterministic control: If every single vehicle had to pre-register its trip, an intelligent central controller could plan the most efficient utilization of the roads and schedule the trips accordingly. This method would result in the optimal use of the infrastructure, avoiding traffic jams and providing deterministic prediction of all arrival times. However, it does come at the expense of enormous complexity. We should also note that this approach doesn’t solve the overload issue. Some trips simply cannot be made, and those requests are rejected by the controller.
The 802.1 time-sensitive networking toolbox provides a rich set of methods that can be selected and adapted to any need.
Making packet networks time-sensitive
The methods outlined above are also applicable to packet networks. The IEEE 802.1 time-sensitive networking (TSN) group specifies a broad range of technologies to connect critical applications over packet networks. It describes device- and network-based methods to prevent congestion for time-critical packets by intelligent scheduling and policing, resource reservation, admission control, accurate synchronization, and end-to-end control.
A combination of the following essential technologies assures timely packet transport even under unfavorable conditions:
- Specific packet forwarding mechanisms to prioritize time-sensitive traffic
- Higher interface capacity to send packets quicker
- A method to deliver very precise time
- A well-designed and managed network, preferably with real-time control of resource utilization
Those different device and network capabilities are well addressed by various IEEE work groups and related specifications. Several methods can improve IEEE 802.1Q Ethernet bridging by advanced scheduling mechanisms and introducing TDM methods with timeslots reserved for express traffic.
Please note that the introduction of a TDM frame structure to Ethernet will likely require additional network-wide synchronization to also align with this TDM structure in each intermediate transport node, assuring priority for express traffic. The frame pre-emption approach works without a need for fixed-time slots as low-priority traffic can be truncated even in the middle of a packet to immediately free up resources for the outgoing time-sensitive data. While this approach creates minor overhead, it can easily interoperate with standard Ethernet interface technologies.
The 802.1 TSN toolbox provides a rich set of methods that can be selected and adapted to any need. Profiles are defined with methods and related configurations to meet application-specific requirements of a time-sensitive packet network. Such profiles have been defined for, like fronthaul in mobile networks (IEEE 802.1CM), audio-video bridging, and industrial automation.
ADVA’s approach to time-sensitive networking
Low delay and minimum delay variations are addressed in several ways by ADVA’s solution portfolio.
Increasing interface speeds are a very efficient way to reduce delay as long frames are sent faster, reducing the queueing time of express traffic. A 1500-byte frame creates a delay of roughly 120ns at a 100Gbit/s interface. ADVA is continuously pushing the limits of Ethernet networking and especially high-speed Ethernet demarcation and aggregation. We launched the first MEF-compliant, ultra-compact 100G demarcation device in 2018. What’s more, with the development of low cost 100Gbit/s interfaces, ADVA is enabling low-latency, high-bandwidth networking.
For many years, ADVA has been researching methods for time-sensitive networks. An innovative aggregation method based on frame gap preservation won the award for best showcase demo at ECOC 2018. And at the upcoming ECOC 2020 in December, a patent-pending delay-correction method will be introduced on top of preemption for eliminating any residual delay variations.
While synchronization is delivered with time-stamped packets in the data plane using IEEE 1588 PTP, this traffic should be handled differently in packet forwarding devices. The timestamps need to be processed by hardware, and packet delay needs to be measured and compensated. Such transparent/boundary clock capabilities assure a minimum time error of less than 5ns caused by the forwarding device. Oscilloquartz, a division of ADVA, is known for providing these capabilities with the most accurate and robust synchronization networks.
What’s more, ADVA’s intelligent end-to-end network control is a very efficient way to counter network congestion by resource planning and traffic policing across the entire network.
In summary, TSN is not a device capability; it’s a network characteristic. It builds on devices with specific features for data forwarding but also handling of PTP in combination with an end-to-end planning and control. At ADVA, we can be your partner in designing and deploying time-sensitive networking.