Skip to main content

IEEE 802.1Qch CQF: cyclic queuing and forwarding, one hop per cycle

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

Cyclic Queuing and Forwarding (CQF) makes latency independent of the other traffic. A frame advances exactly one hop per cycle, so its delay depends only on the path length. No new hardware is needed: CQF is a particular way of configuring two TSN tools that already exist.

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 CQF acts: at the ingress, where a stream gate (802.1Qci) chooses the queue, and at transmission selection, where a transmission gate (802.1Qbv) chooses which queue may send, both driven by the common 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 (8)
AbbreviationStands forMeaning
ATSAsynchronous Traffic ShapingIEEE 802.1Qcr: per-stream token-bucket shaping at every hop using local clocks only.
CBSCredit-Based ShaperThe IEEE 802.1Qav shaper that limits a traffic class to its reserved rate (idleSlope) with a per-queue credit. In MEF bandwidth profiles, CBS instead means Committed Burst Size.
CQFCyclic Queuing and ForwardingIEEE 802.1Qch: two queues per class alternate every cycle, so a frame advances one hop per cycle.
IEEEInstitute of Electrical and Electronics EngineersThe professional association whose Standards Association publishes the IEEE 802 LAN/MAN standards.
IPVInternal Priority ValueA priority assigned inside a bridge (for example by a PSFP stream gate) that selects the traffic class, independently of the frame’s PCP.
PSFPPer-Stream Filtering and PolicingIEEE 802.1Qci: stream filters, stream gates and flow meters that enforce each stream’s contract at ingress.
TASTime-Aware ShaperIEEE 802.1Qbv scheduled traffic: a gate per queue, opened and closed by a repeating gate control list.
TSNTime-Sensitive NetworkingThe set of IEEE 802.1 standards that give Ethernet bounded latency, time synchronisation and high reliability.

1. Definition​

Cyclic Queuing and Forwarding, IEEE 802.1Qch (Annex T of IEEE 802.1Q), divides time into cycles of equal length TT, synchronised across the network. Two queues serve one traffic class and swap roles every cycle:

  • during a cycle, frames received by a bridge are stored in one queue;
  • the frames received in the previous cycle are sent from the other queue.

A frame received in one cycle is therefore sent in the next [2][3].

Cyclic queuing timing on four links: each frame stays one cycle at each hop
Figure 2. CQF timing over four links: each frame of a stream stays exactly one cycle at each hop.Source: J. Farkas, “Introduction to IEEE 802.1: Focus on the Time-Sensitive Networking Task Group,” IEEE 802.1, 15 May 2017, slide 36; cropped, otherwise unchanged.

2. How it is built from gates​

Cyclic queuing and forwarding as PSFP and TAS gates in two switches
Figure 3. CQF as a joint configuration of PSFP and TAS, one time slot: in each switch, PSFP assigns arriving frames to queue 7 while TAS transmits from queue 6.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. 12; cropped, otherwise unchanged.

Standardisation-wise, CQF is not a separate shaper. It is a joint configuration of Per-Stream Filtering and Policing and the Time-Aware Shaper [4]:

  1. The stream gate of PSFP sets the internal priority value (IPV) of arriving frames, alternating between two values each slot, for example 7 and 6.
  2. The transmission gates of TAS run out of phase: while arriving frames go to queue 7, gate 7 is closed and gate 6 is open, and vice versa.

When bridge A opens a gate and sends, bridge B classifies those frames into its own closed queue. That queue opens exactly TT later. Both gating mechanisms need synchronised time, and every device carries the same configuration [4].

Cyclic queuing: time-aware ingress filters feeding two class queues with cycle gates
Figure 3b. The task group’s drawing of CQF: time-aware ingress filters at every input port steer frames into two class queues, each followed by a cycle gate.Source: J. Farkas, “Introduction to IEEE 802.1: Focus on the Time-Sensitive Networking Task Group,” IEEE 802.1, 15 May 2017, slide 35; cropped, with neighbouring slide text, logos or navigation icons blanked.

3. The cycle-time condition​

CQF works only if every frame sent in a cycle reaches the next bridge and is queued before that cycle ends. The cycle must therefore cover:

  • the transmission time of all frames the bridge may send in one cycle, which is what admission control limits;
  • the propagation delay of the link;
  • the processing delay of the receiving bridge;
  • the synchronisation error between the two bridges.

The tutorial on decentralised latency models expresses this as a latency threshold per hop, Δ≤T−dmisc\Delta \le T - d_{\text{misc}}. Here dmiscd_{\text{misc}} accounts for synchronisation inaccuracy. If admission control keeps the worst-case delay of the hop below Δ\Delta, the queue is guaranteed to be empty at the end of its cycle. The downstream shaper then sees every frame as eligible exactly TT after the upstream one [4].

End to end, the latency is about TT per hop. It varies by at most about one cycle, depending on where in a cycle the frame first arrives. This is why CQF gives very low jitter and why its bound needs no knowledge of the other flows [2].

4. Costs and comparison​

The same tutorial evaluates CQF against strict priority, ATS and CBS, using latency models that every bridge can compute locally [4]:

  • Half the traffic classes. Each class needs two queues, so with eight queues only four classes remain for reserved traffic. Priorities are merged, for example 7 and 6 into 7, which increases interference within a class.
  • Best jitter under a global configuration. When one setting is applied network-wide, CQF minimised jitter.
  • Beaten by per-hop tuning. When latency thresholds could be optimised hop by hop, strict priority with a network calculus bound gave better latency and jitter than CQF.
  • A cycle is a hard minimum. Even a lightly loaded path pays TT per hop, so TT trades bandwidth per cycle against latency.

For network calculus, CQF is the simplest case of all. Each hop is a delay element of TT with no burst growth. The burstiness cascade that complicates aggregate multiplexing does not occur.

5. Summary​

PropertyCyclic Queuing and Forwarding
StandardIEEE 802.1Qch, Annex T of IEEE 802.1Q
Built fromStream gates (802.1Qci) and transmission gates (802.1Qbv)
Queues per classTwo, alternating every cycle
Needs time synchronisationYes
LatencyAbout one cycle TT per hop
Main costHalf the traffic classes; a full cycle per hop even at low load

References​

  1. IEEE 802.1, Time-Sensitive Networking (TSN) Task Group. Web page
  2. Wikipedia, "Time-Sensitive Networking," section on cyclic queuing and forwarding. Article
  3. J. Farkas, "Introduction to IEEE 802.1: Focus on the Time-Sensitive Networking Task Group," 15 May 2017, slides 35–36. 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 VI-C and VII-F. arXiv