Markdown Version | Transcript | Session Recording | Session Materials
Session Date/Time: 17 Mar 2026 01:00
ONSEN
Summary
The ONSEN (Operationalizing Network and Service Abstractions) Birds of a Feather (BoF) session was held at IETF 125 to discuss the formation of a new working group focused on the practical deployment, maintenance, and integration of IETF-defined network and service data models. The proponents identified significant gaps in how existing models (such as L2SM, L3SM, L2NM, and L3NM) are operationalized in multi-vendor and multi-domain environments, specifically regarding life-cycle management, service assurance integration, and alignment with OSS/BSS frameworks like TMF APIs. Following presentations on the problem statement, proposed work items, and a draft charter, a series of polls indicated strong community support for forming a working group.
Key Discussion Points
Problem Statement
Benoit Claise presented the ONSEN Problem Statement.
- Historical Context: Referencing the legacy of RFC 3535 and the recent IAB NEMOPS (Next Era of Network Management Operations) workshop, it was noted that while the IETF has produced many models, operational barriers prevent widespread implementation.
- Current Gaps: Existing models like L3SM (Layer 3 Service Model) and L3NM (Layer 3 Network Model) often lack support for dynamic workflows, Service Level Objectives (SLOs), and temporal concepts (e.g., on-demand bandwidth).
- Duplication and Integration: There is significant overlap between models (e.g., L2SM and L3SM), forcing operators to maintain redundant data. Furthermore, integrating YANG-based IETF models with TMF-based OSS/BSS APIs remains a high-friction manual process.
- Life-cycle and Assurance: Standardized mechanisms for service monitoring, rollback, and closed-loop automation are fragmented or missing.
Discussion:
- Qin Wu questioned the extent of duplication, suggesting that L2 and L3 models were intentionally separate to address technology-specific needs.
- Thomas Graf emphasized that mapping TMF models to IETF YANG models frequently fails without vendor-specific "glue," which ONSEN should aim to standardize.
- Italo Busi cautioned against creating "N+1" incompatible models and urged the group to focus on rationalizing existing work.
- Adrian Farrel suggested framing the problem in terms of positive goals (what to achieve) rather than just criticizing existing models like L3SM.
Operationalizing TMF & YANG based APIs
Brad (remote) presented Operationalising TMF & YANG based APIs.
- Business vs. Network Layers: OSS vendors typically use TMF APIs for business-layer interactions, while the network relies on YANG. The mismatch in data models leads to high integration costs for operators.
- Semantic Shift: TMF is moving toward autonomous networks and declarative/semantic interfaces. The IETF needs to ensure its models can be consumed in this new paradigm without requiring prescriptive, manual mapping.
Review of Proposed Work
Chongfeng Xie presented the Review of Proposed Work, outlining five core work items:
- Operational Motivation: Documenting use cases and why current models are difficult to use in production.
- Network Model Updates: Supplemental work for L2NM and L3NM (e.g., SRv6 support, BFD parameters).
- Service Model Updates: Enhancing L2SM and L3SM to reduce duplication and add missing features (e.g., L2 QoS).
- Reusable Framework: Developing building blocks (like site definitions) that can be shared across different services.
- OSS/BSS Interface: Defining mechanisms to map YANG schemas to external APIs like TMF or OpenAPI.
Discussion:
- Wim Henderickx advocated for a focus on "composability," arguing that connectivity should be the primary abstraction, with L2/L3 as properties rather than separate rigid models.
- Daniele Ceccarelli raised concerns regarding backward compatibility for models already in production, such as L2NM and L3NM.
- Kent Watsen noted that a data model transformation framework (mapping service-level models to device-level models) was a key takeaway from the NEMOPS workshop and should be considered.
- Kris Lambrechts highlighted that site definitions are the most critical reusable component for customer-facing interfaces.
Proposed Charter Discussion
Dhruv Dhody and Joe Clarke presented the Proposed Charter Discussion.
- The charter positions ONSEN as a focal point for coordinating service and network abstractions across the IETF, focusing on maintenance, refactoring, and reusable components.
- Relationships: The group will collaborate with TEAS, BESS, OPSAWG, and NETMOD. Technology-specific modeling remains in those respective groups, while generic abstractions move to ONSEN.
- Terminology Clarification: Kent Watsen and Adrian Farrel pointed out that the term "YANG API" in the charter was confusing, as YANG is a modeling language. Proponents agreed to clarify that the work involves the interfaces/APIs derived from those models.
Decisions and Action Items
- Polls:
- Is the problem statement clear and solvable? Overwhelming support (Strong Yes).
- Is there a community willing to author and review? Strong support; several operators (China Telecom, China Unicom, Deutsche Telekom, Telefonica) committed resources.
- Is a new working group required? Strong consensus for a new WG.
- Is there support for the proposed charter? Strong support, with the understanding that specific edits (e.g., regarding "YANG APIs" and backward compatibility) will be addressed.
- Action Item: Proponents to refine the charter on GitHub based on feedback from Wim Henderickx, Kent Watsen, and Adrian Farrel.
Next Steps
- The AD (Mahesh Jethanandani) and the IESG will review the BoF results to decide on working group formation.
- The proponents will continue to iterate on the charter text in the ONSEN GitHub repository.
- Interested participants are encouraged to join the
onsenmailing list (formerlyonions) to discuss draft adoptions and technical gaps.