DETNET

Summary

The DetNet Working Group (WG) met at IETF 126. The session covered the status of the WG, progress of active WG documents, updates on newly adopted drafts, and several individual submissions. A key focus of the meeting was the technical alignment and coordination with ITU-T Study Group 13 (SG13) on deterministic metadata and stateless fair queuing.

Key Discussion Points

WG Status and Liaison Statements

Presentation: 01) Intro, WG Status, Draft Status
Presented by: János Farkas, Lou Berger
Discussion:
- Two new RFCs have been published from the RAW (Reliable and Available Wireless) work: RAW Technologies and RAW Architecture.
- The WG received two liaisons from ITU-T SG13. The most recent one (for action) requests feedback on an ongoing draft recommendation regarding scalable stateless service rate estimation for admission control, which has close relations to draft-ietf-detnet-stateless-fair-queuing.
- Lou Berger noted that the WG needs to coordinate a response to ITU-T SG13 and invited contributions from the WG.

Dataplane Enhancement Taxonomy

Presentation: 02) Dataplane Enhancement Taxonomy
Draft: draft-ietf-detnet-dataplane-taxonomy
Presented by: Jeong-dong Ryoo
Discussion:
- Section 8 (Considerations for Interoperability between Solutions) has been revised to reference draft-ietf-detnet-multi-domain-framework. Interoperability levels are analyzed on the control plane (operability) and data plane (performance metrics).
- Mike Johnston asked how an implementer should interpret "low, medium, or high" interoperability levels. Jeong-dong Ryoo responded that this is an informational document aimed at categorization rather than providing specific implementation specifications.
- Lou Berger invited further contributions on the mailing list to improve the interoperability definitions.

Latency Guarantee with Stateless Fair Queuing

Presentation: 03) Latency Guarantee with Stateless Fair Queuing
Draft: draft-ietf-detnet-stateless-fair-queuing
Presented by: Jeong-dong Ryoo
Discussion:
- Revision 01 adds the service rate estimation process in Section 6.3, referencing the ITU-T SG13 work item Y.IETF-QS-SRE-RF. An additional metadata field ("difference of finish times") has been introduced, bringing the metadata count to four, and corresponding header formats for MPLS (MNA post-stack) and IPv6 were added.
- Lou Berger expressed concern over potential duplication of work between IETF and ITU-T. He emphasized that the IETF document must reference, not duplicate, the ITU-T definitions to avoid being blocked during the IESG review.
- The responsible Area Director (AD) strongly recommended resolving any overlapping or complementary scope issues within the WG rather than waiting for the IESG stage.
- Mike Johnston noted that the addition of the fourth metadata field resolves a mismatch with one ITU-T draft but introduces inconsistencies with two already completed ITU-T recommendations (Y.3129 and Y.3148), which only use three metadata fields.

Deadline Based Deterministic Forwarding

Presentation: 06) Deadline Based Deterministic Forwarding
Draft: draft-ietf-detnet-deadline-based-forwarding
Presented by: Peng Liu
Discussion:
- Revision 01 simplifies the document and focuses on two optimal stateless options for scalability: "latency compensation + cycle queueing (CQ)" and "latency compensation + Rotating Priority Queue (RPQ)". Options requiring per-flow state on core nodes have been removed.

Timeslot Queueing and Forwarding Mechanism

Presentation: 07) Timeslot Queueing and Forwarding Mechanism
Draft: draft-ietf-detnet-packet-timeslot-mechanism
Presented by: Peng Liu
Discussion:
- Revision 01 adds operational considerations, moves the evaluations to the appendix, removes PTM detection, and removes the global time slot.
- Jeong-dong Ryoo asked if the authors had considered merging this draft with Tagged Cyclic Queuing and Forwarding (TCQF) (draft-ietf-detnet-tcqf), pointing out that TCQF is class-based while this draft is flow-based. Xiaofu Ji added that there are no technical blockers for a merger, but it depends on the WG's decision.

On-time Forwarding with Push-In First-Out (PIFO) Queue

Presentation: 08) On-time Forwarding with Push-In First-Out (PIFO) queue
Draft: draft-ietf-detnet-ontime-forwarding
Presented by: Yunchao Liu
Discussion:
- Revision 02 introduces Appendix A, providing mathematical proof for minimum and maximum end-to-end latency bounds under configured admission conditions. It also mentions an FPGA implementation of the PIFO queue scheduler.

On-time Forwarding with Non-Work Conserving Stateless Core Fair Queuing

Presentation: 10) On-time Forwarding with Non-Work Conserving Stateless Core Fair Queuing
Draft: draft-ietf-detnet-nscore
Presented by: Yunchao Liu
Discussion:
- Revision 02 introduces an "arrival-based" eligibility time (ET) and finish time (FT) calculation method (Section 5.2.2). This method uses local arrival time, scheduling gap, and a delay component in the packet header, eliminating the need for strict synchronization or propagation delay awareness.
- The solution has been implemented on FPGA and is currently undergoing performance testing.

A Control Plane Framework for Multi-Domain Deterministic Networking (DetNet)

Presentation: 11) A Control Plane Framework for Multi-Domain Deterministic Networking (DetNet)
Draft: draft-ietf-detnet-multi-domain-framework
Presented by: Carlos J. Bernardos
Discussion:
- The draft has been adopted by the WG. Revision 01 incorporates feedback from the adoption call, adding terminology for multi-domain admission control and cycle alignment, plus requirements for coordinated resource reservation and cycle-phase alignment at boundaries.
- Jeong-dong Ryoo suggested that the terminology in Section 2 should reflect that time-bounded queueing solutions are not always slotted (e.g., in-time EDF).
- Lou Berger (as a contributor) cautioned against redefining existing terms such as "admission control" and "resource reservation" which are already defined in TE and DetNet architectures. He recommended referencing the existing specifications or using verbatim copies to maintain consistency.
- Mike Johnston sought clarification on the definition of "multi-domain." Carlos J. Bernardos clarified that the draft treats a domain as the scope controlled by a single controller (such as a PCE), which might reside within a single administrative entity.

Deterministic Networking SRv6 Data Plane for Service Protection

Presentation: 12) Deterministic Networking SRv6 Data Plane for Service Protection
Draft: draft-ietf-detnet-srv6-svc-protection
Presented by: Presenter from Ericsson (on behalf of the authors)
Discussion:
- The draft has been adopted by the WG. The authors presented two independent implementations: the "Dependable Networking Toolkit" (a portable user-space tool supporting Layer 2/Layer 3 DetNet) and "XDP FRER" (using Linux eXpress Data Path for high-performance service protection).
- Future work includes detailing redundancy policy behavior in DetNet and clarifying interworking with existing SRv6 L2/L3 VPN deployments.

Flow Aggregation for Enhanced DetNet

Presentation: 13) Flow Aggregation for Enhanced DetNet
Draft: Individual submission
Presented by: Presenter
Discussion:
- The draft title was changed to focus on a "framework" for flow aggregation. Section changes clarify the motivation, explain the relationship between aggregate and member flows (using C-Score/N-Score examples), and outline data plane options (including MPLS MNA).
- János Farkas questioned the use of BGP Flowspec for mapping DetNet flows to TSN streams in a cross-domain scenario. The presenter replied that BGP Flowspec is one proposed control plane alternative for handling boundary mapping.
- Lou Berger (as a contributor) suggested checking the DetNet YANG model, which already includes support for aggregation, to ensure alignment at both the data plane and configuration/control plane levels.

Mechanism to Control Jitter Caused by Policing in DetNet

Presentation: 14) Mechanism to control jitter caused by policing in Detnet
Draft: Individual submission
Presented by: Xiaofu Ji
Discussion:
- The terms "policing delay" and "policing delay budget" were changed to "shaping delay" and "shaping delay budget" to avoid confusion. The draft describes the relationship between ingress shaping delay, egress dumping delay, and the overall path delay budget.
- Jeong-dong Ryoo questioned how the system guarantees that path delay remains within the overall latency bounds and pointed out that the proposed mechanism overlaps significantly with existing standards (such as ITU-T ATM-related or similar standards). He advised the authors to review these existing mechanisms.

Handling of Traffic Characteristic Deviation for DetNet

Presentation: 15) Handling of traffic characteristic deviation for DetNet
Draft: Individual submission
Presented by: Presenter from China Unicom
Discussion:
- The term "anomalous traffic" was replaced with "traffic characteristic deviation" to avoid misleading readers. The draft introduces "squeezing" (deferring packets to the next slot) and "degrading" (downgrading priority and modifying schedule identifiers) to handle burst accumulation.
- Jeong-dong Ryoo noted that burst accumulation is typically a network-induced issue rather than the fault of an individual flow. He cautioned that squeezing or degrading might penalize conforming, innocent flows, and stressed the importance of flow isolation.
- Lou Berger (as a contributor) questioned whether "squeezing" is technically distinct from "shaping" applied to a time-slotted queue, or if it is merely a new term for an established concept. He suggested continuing this discussion on the mailing list.

PREOF Observability and OAM Requirements

Presentation: 16) PREOF Observability and OAM Requirements
Draft: Individual submission
Presented by: Luis Miguel Contreras
Discussion:
- This new draft defines specific OAM requirements, metrics, and observable events (replication, elimination, reordering, timeouts, late arrivals) for Packet Replication, Elimination, and Ordering Functions (PREOF).
- Balázs Varga noted that many of the proposed metrics might be implementation- or algorithm-specific (for instance, identifying which path a packet arrived on is not always possible on some data planes). He also requested clarifying how this draft relates to the existing DetNet OAM document.
- Lou Berger agreed that some metrics (e.g., tracking sequence numbers of all discarded packets) are challenging to implement at high data rates and advised dividing the requirements into "required" versus "desired/optional" metrics.
- Jeong-dong Ryoo asked for clarification on the difference between "packet discard" and "elimination" in Requirement 3. Luis Miguel Contreras agreed to refine the definition.
- Balázs Varga suggested looking at open-source PREOF implementations (and IEEE 802.1CB) as they already implement some of these observability counters.

Joint Coordination with ITU-T SG13

Discussion:
- Taesang Choi (Rapporteur for ITU-T SG13 Question 6) attended the session and confirmed that ITU-T SG13 focuses on high-level use cases, requirements, and framework/architecture. He welcomed a clear division of work where DetNet specifies the actual protocols and data plane formats. He noted that the ITU-T draft recommendation was delayed specifically to align with IETF DetNet.
- Lou Berger suggested that the IETF response to the ITU-T liaison should explicitly ask for clarification on the division of work.
- Mike Johnston reiterated that the WG must resolve the metadata count inconsistency (3 vs. 4 metadata fields) to ensure interoperability between the systems defined by both organizations.

Decisions and Action Items

Action Item: The WG will draft a response liaison to ITU-T SG13 addressing their request for feedback, aiming to clarify the boundary of work between IETF data plane specifications and ITU-T framework requirements.
Decision: Authors of draft-ietf-detnet-stateless-fair-queuing must ensure that all metadata and header format definitions are fully aligned with the final ITU-T specification to prevent blocking at the IESG.

Next Steps

Detailed mailing list discussion is expected for the newly adopted control plane framework (draft-ietf-detnet-multi-domain-framework) and the newly adopted SRv6 service protection draft (draft-ietf-detnet-srv6-svc-protection).
Authors of individual drafts regarding jitter control, traffic deviations, and PREOF OAM will refine their documents based on the feedback received during the session.

Automatic IETF Minutes