Time-Sensitive Networking (TSN) is a set of standards designed to enable deterministic communication for ethernet based networks. TSN offers different tools for different deterministic network requirements.
Schedule traffic (IEEE 802.1Qbv) and frame preemption (IEEE 802.1Qbu & 802.3br) are some of the most appropriate TSN tools available to achieve ultra-low latency over multiple hops. Schedule traffic works by allocating dedicated time slots for different types of traffic, similar to a Time-Division Multiple Access (TDMA) scheme. However, configuring the dedicated time slots at each network hop in a way that guarantees no waiting time for frames traversing the network is easier said than done.
Riding the “green wave” in an ethernet network
When driving a car on city streets it is always nice to experience a green wave. A green wave occurs when a series of traffic lights (usually three or more) are coordinated to allow continuous traffic flow over several intersections in one main direction. The traffic lights need to be configured based on the distance between each other and the expected cars’ speed.
Similarly, for an ethernet network, a scheduled traffic green wave occurs when a series of bridges and end stations are configured in a coordinated way to allow continuous traffic flow for the scheduled streams over several hops. Cable and bridges delays as well as link speeds need to be taken into account when configuring the network.
Each Qbv enabled transmit port on a bridge or end station has a gate parameter table. The gate parameter table is characterized by:
The base time: The time when the schedule starts
The cycle time: The time after which the control list index is returned to zero, and
The control list: An ordered lists where each entry tells which queue is open or closed and for how long
Configuring the schedule traffic quickly while assuring it works as expected
One easy solution to configure schedule traffic over multiple hops to achieve a green wave is to use the base time to compensate for cables and bridges delays. For example, for the simple network topology presented above, we can start with base time 0.0 at talker 1.
On bridge 1, the base time should be the cable delay between talker 1 and bridge 1 plus the time it takes the frame to traverse bridge 1. In case of a store and forward bridge, the frame length also needs to be considered. For cut-through bridges this can be ignored.
To generalize this formula, at hop n, the base time should be set to:
The same approach also works for more complex network topologies as well as for multiple streams with multiple talkers and listeners. Cables and bridges delays can be calculated dynamically. The network configuration can be updated in real time by the CNC and CUC to maintain a green wave when new streams are joining the network or when traffic conditions change.
Measuring end-2-end latency is not enough
To assure the desired ultra-low latency is achieved, measuring end-2-end latency is not enough. Metrics for assessing the stability of scheduled streams over time and over multiple hops need to be collected at different points in the network. The most important metrics to evaluate schedule traffic are:
Deviation from expected arrival time for each frame part of a given scheduled stream
• Max positive deviation
• Max negative deviation
• Average deviation
Scheduled traffic timed histograms – e.g. how many frames are received within:
• 0..100ns from expected time
• 100..500ns from expected time
• 0.5..2us from expected time
Ready to start using TSN in your systems? Learn more about how Spirent can help you realize the