Skip to main content

IEEE 802.1Qcr ATS: asynchronous traffic shaping without a clock

· 7 min read
PhD candidate, LyRIDS (ECE) × CEDRIC (Cnam)

Class-based shapers let the bursts of a stream grow from hop to hop, and scheduled traffic needs a network-wide clock and a central schedule. Asynchronous Traffic Shaping (ATS) avoids both problems: it restores each stream's original traffic contract at every hop, using only the bridge's local clock.

The IEEE 802.1Q bridge: architecture and data-plane actionsIEEE 802.1Q bridge ↗Control planeSRP (Qat), TSN configuration (Qcc), path control (Qca)frame inframe outIngressPSFP (Qci)FRER (CB)Relayforwardingexplicit treesQueuingCBS (Qav)ATS (Qcr)Tx selectiongates (Qbv)CQF (Qch)MAC / PHYpreemption(Qbu, 802.3br)Common time: gPTP (802.1AS)
Figure 1. Where ATS acts: per-stream state is selected at the ingress, by the same stream filter as PSFP, and applied in front of shared shaped queues at the egress. It does not touch the time base.Source: own drawing, after the bridge architecture, data-plane actions and QoS function chain in J. Farkas, TSN Basic Concepts, DetNet–TSN workshop, 11 November 2018, slides 7, 8 and 10.
Abbreviations used in this post (5)
AbbreviationStands forMeaning
ATSAsynchronous Traffic ShapingIEEE 802.1Qcr: per-stream token-bucket shaping at every hop using local clocks only.
BEBest EffortTraffic with no guarantee of delay or delivery.
FRERFrame Replication and Elimination for ReliabilityIEEE 802.1CB: frames are numbered, sent over disjoint paths, and duplicates are removed.
IEEEInstitute of Electrical and Electronics EngineersThe professional association whose Standards Association publishes the IEEE 802 LAN/MAN standards.
PSFPPer-Stream Filtering and PolicingIEEE 802.1Qci: stream filters, stream gates and flow meters that enforce each stream’s contract at ingress.

1. Definition​

Asynchronous Traffic Shaping, IEEE 802.1Qcr, is based on the urgency-based scheduler of Specht and Samii. Each stream gets its own token bucket at every bridge, with a committed burst size bb and a committed rate rr. A frame becomes eligible for transmission at the earliest time its bucket holds enough tokens for it. Strict priority then serves eligible frames [2][3][4].

Two features distinguish it from plain per-flow shaping:

  • Shared queues. Streams share a small number of shaped queues. The shaper state of each frame's stream is applied to the frame at the head of the queue [3].
  • No synchronisation. Eligibility times are computed from the local clock. A frame whose eligibility time lies too far in the future, beyond a configured maximum residence time, is discarded [2].
ATS chain: filter, gate with internal priority, ATS shaper, then queuing
Figure 2. The ATS components on the original slide: a filter, a gate that sets the internal priority, the ATS shaper, then queuing.Source: J. Farkas, “TSN Basic Concepts,” DetNet–TSN workshop, IEEE 802.1, 11 November 2018, slide 15; cropped, with neighbouring slide text, logos or navigation icons blanked.
Asynchronous traffic shaping per hop: high and low priority shaped queues on both sides of a link
Figure 2b. ATS hop by hop: traffic is reshaped in high- and low-priority shaped queues at every hop, so the pattern after each link stays smooth even when best-effort (BE) traffic shares the link.Source: J. Farkas, “TSN Basic Concepts,” DetNet–TSN workshop, IEEE 802.1, 11 November 2018, slide 14; cropped, otherwise unchanged.

2. Queue allocation rules​

Sharing a queue between streams can cause head-of-line blocking: a frame that is not yet eligible holds up an eligible frame behind it. ATS limits this with queue allocation rules. Frames go to separate shaped queues if they differ in [4]:

  • the ingress port they arrived on (the upstream transmitter);
  • the egress port;
  • the traffic class in the upstream bridge or in the current bridge.

Inside one queue, all frames then come from the same upstream queue, so they arrive in an order compatible with their eligibility times.

3. Delay segments: from shaper to shaper​

Delay segments between shapers of consecutive switches
Figure 3. Shaper-to-shaper delay segments, the borders of responsibility for a local bound: each segment runs from the shaper of one switch to the shaper of the next.Source: A. Grigorjew, S. Geissler, J. Schmitt, T. Hossfeld, “Decentralized Latency Models and Information Flow for TSN Shapers – A Tutorial,” arXiv:2610.11999, 2026, Fig. 5; cropped, otherwise unchanged.

The ATS analysis does not measure delay from ingress to ingress. It measures from the end of one shaper to the end of the next, as in Figure 3. This boundary was first used in the ATS latency analysis of Specht and Samii. It was proposed for the P802.1DD Resource Allocation Protocol, and a 2026 tutorial applies it to all shapers [4].

The choice matters because traffic leaving a token-bucket shaper is known to conform to (b,r)(b, r). Each segment therefore starts from a clean, well-defined arrival curve, whatever happened upstream. The downstream bridge computes the bound for the segment that ends in it, using what its neighbour tells it [4]. The stream reservation post lists the information it receives.

Switch delay model: propagation, store and forward, processing, shaper queuing and priority queuing
Figure 3b. The switch delay model of the tutorial: propagation, store and forward, processing, shaper queuing and priority queuing between the RX and TX ports.Source: A. Grigorjew, S. Geissler, J. Schmitt, T. Hossfeld, “Decentralized Latency Models and Information Flow for TSN Shapers – A Tutorial,” arXiv:2610.11999, 2026, Fig. 4; cropped, otherwise unchanged.

4. Why the bound does not cascade​

With class-based shapers or plain strict priority, a stream's burst grows by its rate times the delay variation it has accumulated:

beff=b+r (L+−L−),b^{\text{eff}} = b + r\,(L^{+} - L^{-}),

where L+L^{+} and L−L^{-} are its accumulated maximum and minimum latencies (the fundamental bounds post gives the general result). ATS removes the extra term at every hop. Its per-segment bound uses the original burst bb declared by the talker. It also needs no pre-configured latency threshold, although its queue allocation rules require more queues [4].

In network calculus, the ATS shaper is an interleaved regulator. Its key property, shown by Le Boudec, is that an interleaved regulator placed after a FIFO system does not increase that system's worst-case delay. Re-shaping is therefore "for free" in the worst case [5]. A review of industrial IoT networking cites the same property: ATS does not increase the worst-case delay through the network [6].

5. How ATS compares​

The 2026 tutorial compares five decentralised latency models on industrial and linear topologies [4]:

  • ATS performed well everywhere. It gave the best burstiness at the receiver and worst-case latency similar to the best strict-priority model.
  • Strict priority with a network calculus bound can still win at the highest priority, especially in bridges with very few ports, because frames from one ingress link cannot arrive faster than that link (serialisation). At lower priorities ATS consistently did better.
  • The main disadvantage is hardware. Per-stream state and extra queues make ATS costly and still rarely available.

ATS is also a remedy for a redundancy side effect. After FRER eliminates duplicates, the merged stream can be burstier than the original. Placing ATS at the merge point reshapes it, although combining FRER and ATS is still being standardised [4].

The industrial review summarises the division of labour. ATS suits sporadic or aperiodic critical traffic, such as alarms and events. The credit-based shaper suits regular but bursty traffic, and the time-aware shaper suits regular critical traffic [6].

6. Summary​

PropertyAsynchronous Traffic Shaping
StandardIEEE 802.1Qcr
Acts onPer-stream state at the ingress, shared shaped queues at the egress
StateOne token bucket (b,r)(b, r) per stream per bridge
Needs time synchronisationNo; local clocks only
BoundUses the talker's original burst; no burstiness cascade
Main costPer-stream state, more queues, limited hardware support

References​

  1. IEEE 802.1, Time-Sensitive Networking (TSN) Task Group. Web page
  2. Wikipedia, "Time-Sensitive Networking," section on asynchronous traffic shaping. Article
  3. J. Farkas, "TSN Basic Concepts," DetNet–TSN workshop, 11 November 2018, slides 13–15. Slides
  4. A. Grigorjew, S. Geissler, J. Schmitt, T. Hossfeld, "Decentralized Latency Models and Information Flow for TSN Shapers – A Tutorial," arXiv:2610.11999, 2026, Sections I, IV, VI-A, VI-D, VII-F and VIII. arXiv
  5. J.-Y. Le Boudec, "A Theory of Traffic Regulators for Deterministic Networks With Application to Interleaved Regulators," IEEE/ACM Transactions on Networking, vol. 26, no. 6, 2018.
  6. I. Behnke, H. Austad, "Real-Time Performance of Industrial IoT Communication Technologies: A Review," arXiv:2311.08852, 2023, Section III-D. arXiv