Markdown Version | Transcript | Recording 1 | Recording 2 | Session Materials

IPPM

Summary

The IPPM and BMWG working groups held a joint session at IETF 126 in Vienna to discuss active working group drafts, proposed work, and the ongoing effort to merge the two groups under a single joint charter.

Key milestones highlighted during the session include several documents progressing to the RFC Editor queue (Quality of Outcome, Asymmetrical Packets, and various BMWG milestones), updates on active drafts such as connectivity monitoring and responsiveness, and a structured discussion on the joint charter proposal, including implementation policies and potential names for the merged working group.

Key Discussion Points

1. Welcome and Working Group Status Updates

• Presenter: Marcus Ihlar (IPPM Chair) and Giuseppe Fioccola (BMWG Chair)
• Slides: 0. Chair Slides
• BMWG Status Updates:
  • Segment routing benchmarking methodology was submitted to the IESG for publication.
  • The MLR search and YANG model documents are currently in the RFC Editor queue.
• IPPM Status Updates:
  • Quality of Outcome (QoO) and Asymmetrical Packets are in the RFC Editor queue.
  • Hybrid two-step progressed and received AD review; authors are addressing feedback.
  • draft-ietf-ippm-stamp-ext-hdr concluded Working Group Last Call (WGLC); shepherd write-up is pending.
  • draft-ietf-ippm-alt-mark-deployment and draft-ietf-ippm-alt-mark-yang are in WGLC and receiving active feedback.
  • IOAM data integrity and integrity YANG are under IESG evaluation.
• RFC 7799bis Update:
  • Giuseppe Fioccola discussed the effort to transition RFC 7799 to a Best Common Practice (BCP). Based on mixed last-call feedback requesting a refresh, draft-fioccola-ippm-rfc7799bis has been published. A GitHub repository is available for tracking progress.

2. An Experimental Connectivity Monitoring Metric

• Presenter: Rüdiger Geib
• Slides: 1. An Experimental Connectivity Monitoring Metric
• Draft: draft-ietf-ippm-connectivity-monitoring
• Discussion:
  • The draft had been parked, but success criteria for the experiment have now been added to move it forward.
  • The metric uses Segment Routing (SR) to specify a measurement path, creating a specific overlay to evaluate network state (connectivity loss, congestion location) via tomography. It operates on a timescale of seconds to minutes.
  • Success criteria are defined by whether round-trip delay, single link loss, and single node interface congestion can be detected as defined in the document.
  • Tim Chown pointed out that the Datatracker still lists the draft's intended status as "Proposed Standard" rather than "Experimental," which needs to be corrected by the chairs.
  • Marcus Ihlar urged the working group to review the updated document before proceeding to WGLC.

3. Quality of Outcome (QoO) Update

• Presenter: Ike Kunze
• Slides: 2. QoO Update
• Discussion:
  • Although the document is in the RFC Editor queue, the authors want to ensure ongoing interoperability. QoO is highly flexible, meaning implementations may diverge based on differing network measurement techniques and configuration thresholds (latency percentiles, loss rates, throughput).
  • The goal is to share deployment experiences, document common best practices, and define standardized application profiles. A simplified Docker-based emulation dashboard was developed during the hackathon to experiment with different application profiles.
  • Giuseppe Fioccola supported the work, emphasizing the need for granular application profiles (such as for the pro-gaming community).
  • Jason Livinggood indicated that Comcast plans to conduct field testing of QoO and bring operational feedback back to the working group.

4. Responsiveness under Working Conditions (RPM)

• Presenter: Stuart Cheshire
• Slides: 3. Responsiveness under Working Conditions (RPM)
• Draft: draft-ietf-ippm-responsiveness
• Discussion:
  • The draft was updated to version 09. During the hackathon, testing with TCP Prague and L4S initially failed due to a kernel version mismatch on the bottleneck. Once resolved, the L4S behavior was highly stable, maintaining bounded round-trip delay (~21-22 ms) with zero loss or retransmissions.
  • However, the RPM tool reported a discrepancy of 260 ms round-trip time. The team is investigating whether this is a specification issue, an implementation bug, or a fundamental challenge of testing convergence within the tool's quick 10-second test window.
  • Apple is acquiring identical hardware to Greg White's testbed to isolate and resolve the discrepancy.

5. iperf2 Android Client

• Presenter: Stuart Cheshire (on behalf of Bob McMahon)
• Slides: IETF 126 IPPM Bob McMahon iperf2 Android Client
• Discussion:
  • An update on the development of the iperf2 Android client. The tool uses Linux kernel calls to provide a real-time GUI display of the TCP send buffer (including congestion window, bytes in flight, and unsent buffered data).
  • Precompiled APKs are available on SourceForge, and feedback is welcome.

6. In-Network Telemetry Aggregation Using IOAM Aggregation Trace and Template Options

• Presenter: Alexander Clemm
• Slides: 4. In-Network Telemetry Aggregation Using IOAM Aggregation Trace and Template Options
• Discussion:
  • This work explores preprocessing telemetry data (min, max, sum) during packet traversal (in-network aggregation) rather than collecting raw records from each individual node.
  • Hackathon efforts focused on comparing the trace and template options, extending proof-of-concepts (e.g., a network sustainability carbon-tracking case by OST Switzerland and a Linux kernel implementation by the University of Rome), and designing a combined "trace plus template" option that saves 8 bytes of packet overhead.
  • Giuseppe Fioccola and Frank Brockners discussed whether other proposed IOAM extensions in IPPM should align with this template option to prevent option sprawl. Frank Brockners supported working group adoption of the template draft to create a streamlined foundation similar to IPFIX.
  • A poll was taken regarding document familiarity:
    • Poll: Did you read the presented document?
      • Yes: 7, No: 17, No opinion: 2 (Total: 70)

7. AI Fabric Benchmarking Suite

• Presenter: Ching Wu
• Slides: 5. AI Fabric Benchmarking Suite
• Discussion:
  • The benchmarking suite comprises three drafts (Terminology, Training Benchmarking, and Inference Benchmarking), updated to version 03.
  • Changes include consolidating terminology across drafts, adding "packet radar improvement" metrics, adding dynamic flow configurations, and downgrading RFC 2889 to an informative reference.
  • In response to participant questions about testing environments, Ching Wu clarified that the primary focus is laboratory testing, though production-like environments are also being planned.
  • Giuseppe Fioccola noted that this suite is a strong candidate for BMWG adoption.

8. Computing-Aware Traffic Steering (CATS) Benchmarking Methodology

• Presenter: CATS Benchmarking Presenter (presented on behalf of authors)
• Slides: 6. cats-benchmarking-methodology
• Discussion:
  • The draft has reached version 04, incorporating hybrid-mode test setups, refined test-to-metric mappings, and metric aggregation/normalization functions in the appendix.
  • The authors are requesting working group adoption. Giuseppe Fioccola noted that the draft is slated for adoption consideration soon, highlighting the need to gather feedback from CATS WG experts.

Joint Charter Discussion and WG Polls

• Presenter: Ching Wu
• Slides: 7. Feedback on IPPM/BMWG Joint Charter Proposal

Ching Wu presented the proposed joint charter for the combined working group, summarizing feedback received since IETF 125. The draft charter avoids strict separation between production and lab-based applicability, integrates IPFIX and YANG management aspects, and emphasizes calibration and transport scalability.

Implementation Policy

To maintain high-quality specifications, the joint charter proposal evaluated three options for standardizing implementations:

• Option 1: Prioritize documents with existing implementations.
• Option 2: Require at least one implementation for Standards Track documents.
• Option 3: Require at least two implementations for Standards Track documents.
• Outcome: There was general agreement to adopt Option 2.

Joint Working Group Name Polls

A series of polls were conducted to gauge preference for the name of the merged working group:

1. Poll: Please vote "YES" if you favor "BPM" as the name of the joint working group name. If you believe this name should not be chosen, vote "NO" and speak up with the reason why not.
   • Yes: 17, No: 1, No opinion: 14 (Total: 72)
   • Discussion: Tim Chown voted NO, expressing concern that "BPM" sounds like "beats per minute." Jared Mauch questioned what the letters would stand for.
2. Poll: Please vote "YES" if you favor "BIPP" as the name of the joint working group name. If you believe this name should not be chosen, vote "NO" and speak up with the reason why not.
   • Yes: 2, No: 13, No opinion: 8 (Total: 72)
3. Poll: Please vote "YES" if you favor "BMPM" as the name of the joint working group name. If you believe this name should not be chosen, vote "NO" and speak up with the reason why not.
   • Yes: 12, No: 3, No opinion: 6 (Total: 72)
4. Poll: Please vote "YES" if you favor "NETMEASURE" as the name of the joint working group name. If you believe this name should not be chosen, vote "NO" and speak up with the reason why not.
   • Yes: 3, No: 16, No opinion: 4 (Total: 72)
5. Poll: Please vote "YES" if you favor "BPMP" as the name of the joint working group name. If you believe this name should not be chosen, vote "NO" and speak up with the reason why not.
   • Yes: 9, No: 14, No opinion: 4 (Total: 72)

Decision: The name preference will be finalized and confirmed on the mailing list, with "BPM" and "BMPM" showing the highest level of support.

Charter Feedback

• Tim Chown suggested several textual improvements:
  • Soften the phrasing around measurement disruption to recognize that some defined measurements may be temporarily disruptive.
  • Clarify or remove the reference to "security building blocks" to avoid ambiguity.
  • He agreed to submit these as GitHub issues/PRs.

Next Steps

1. Joint Charter:
   • Tim Chown to submit PRs/issues on GitHub to address charter text improvements.
   • The joint charter name and final text will be posted to the mailing list for review before sending it to the IESG.
2. draft-ietf-ippm-connectivity-monitoring:
   • Working group members are requested to review the success criteria added in the latest revision.
   • Chairs will correct the document's intended status in the Datatracker to "Experimental."
3. draft-ietf-ippm-responsiveness:
   • Stuart Cheshire and contributors to investigate the RPM tool's RTT measurement discrepancy and refine the methodology if needed.

Session Date/Time: 23 Jul 2026 07:00

IPPM

Summary

The IPPM working group held a joint session with the BMWG at IETF 126. The meeting covered progress updates on active working group drafts (Alternate Marking, encrypted PDMv2, STAMP ECN, and STAMP MPLS extensions) and presentations on several new benchmarking and performance measurement proposals. These proposals focused on power consumption, source address validation (SAV), RPKI relying party and route origin validation (ROV) behavior, AI fabric switching efficiency, and AI agent operations.

The chairs noted that future meetings might operate under a newly proposed joint charter name, BPM (Benchmarking and Performance Measurement).

Key Discussion Points

1. Welcome and Agenda Bashing

• Presenter: Marcus Ihlar
• Slides: 0. Chair Slides
• Discussion: Marcus Ihlar reviewed IETF policies, administrative details, and the agenda. He noted that the group is currently discussing a joint charter proposal for IPPM and BMWG under a combined name (e.g., BPM). The agenda was approved as proposed.

2. Alternate Marking Deployment Framework & YANG Models

• Presenter: Giuseppe Fioccola
• Drafts: draft-ietf-ippm-alt-mark-deployment and draft-ietf-ippm-alt-mark-yang
• Discussion:
  • Deployment Framework: Working Group Last Call (WGLC) has concluded. Revisions addressed comments from Benoit Claise, Alex Clemm, and Tim Chown. Key updates include distinguishing the base document from complementary work, clarifying on-node vs. centralized collector computation options (referencing RFC 9951/9947), and extending the controller domain definition to cover trust relationships.
  • YANG Models: The configuration model was refined to define profiles under interfaces. The configuration and telemetry models are separated.
  • IPFIX Extension: Refined to improve flow aggregation description and align naming conventions.
  • Next Steps: Move the deployment draft forward. A joint WGLC for the IPFIX draft with the OPSAWG is planned.

3. IPv6 Performance and Diagnostic Metrics Version 2 (PDMv2)

• Presenter: Nalini Elkins
• Draft: draft-ietf-ippm-encrypted-pdmv2
• Discussion:
  • The draft has been revised to remove cryptographic key exchange and security context details from its core scope (suggesting a RADIUS EAP registration flow as an example instead). PDMv2 will focus strictly on the measurements, data structures, and packet contents.
  • PDMv2 operates at the IPv6 layer and needs a new option type allocation from IANA.
  • Feedback: Mohamed Boucadair and Thomas Graf suggested requesting early reviews from the Security, Internet, Operations, and Transport (TSVART) directorates.

4. Power Benchmarking

• Presenter: Shailesh
• Discussion:
  • Updates to the power benchmarking draft include adding strict laboratory environment conditions (temperature, humidity, atmospheric pressure) based on CES 203136.
  • The draft added a "typical power" test scenario and simplified traffic configurations by removing inter-packet delay.
  • Feedback: Giuseppe Fioccola and Shailesh noted that the draft needs reviews from energy efficiency experts (such as the Green WG) before advancing to WGLC.

5. Source Address Validation (SAV) Benchmarking Methodology

• Presenter: Linhui Sun
• Discussion:
  • Clarified intra-domain and inter-domain SAV scopes. Added terms like Limited Propagation Prefix (RPP) and Hidden Prefix (caused by Direct Server Return).
  • Test cases for Customer-Provider interfaces were refined, and the test reporting format was expanded.
  • Lan Chang noted that the cases align with SAVNET working group problem statement drafts.

6. MPLS In-situ OAM (IOAM)

• Presenter: Rakesh Gandhi
• Discussion:
  • Outlined the encapsulation of IOAM options in MPLS using MPLS Network Actions (MNA) in-stack and post-stack. Direct Export (DEX) is supported in both formats.
  • Thomas Graf encouraged the IPPM and BMWG communities to review the draft, as the MPLS working group is preparing for a WGLC.

7. STAMP Class of Service (CoS) Extension for ECN

• Presenter: Greg White
• Draft: draft-ietf-ippm-stamp-cos-ecn
• Discussion:
  • Revision 01 resolves reflector ECN signaling ambiguity by mandating "not ECT" (00) when a reflector cannot set the requested value.
  • Discussed an alternative "CoS v2" encoding proposal that includes explicit DSCP/ECN values for both forward and reverse paths. While this helps on-path observers, it breaks backward compatibility.
  • Giuseppe Fioccola suggested referencing hybrid/on-path measurements (such as draft-ietf-ippm-on-path-active-measurements) to support hop-by-hop measurements.
  • Greg White invited implementations to collaborate on interop testing at the next Hackathon.

8. Performance Measurement (PM) Streaming and Collection

• Presenter: Bin Young
• Discussion:
  • This work was split from a single CCAMP document into three: two generic PM drafts for IPPM and one transport-specific draft remaining in CCAMP.
  • The generic IPPM drafts cover PM collection measurements (including counts, snapshots, and tri-marks based on ITU-T G.7710) and interval capability discovery.
  • Feedback: Thomas Graf proposed initiating a liaison check with ITU-T SG15. Giuseppe Fioccola suggested looking at tagging data based on measurement type (active/passive/hybrid). Chin suggested checking for overlap with NETCONF and LMAP YANG models.

9. RPKI Relying Party Benchmarking

• Presenter: Lan Chang
• Discussion:
  • Focuses on the functional correctness (e.g., certification path validation and syntax checking) of RPKI relying parties (RP), with performance as a secondary objective.
  • Presenters plan to add CRL processing tests and daemon mode operation evaluations in future versions.

10. STAMP Extensions for Reflecting MPLS Headers

• Presenter: Rakesh Gandhi
• Draft: draft-ietf-ippm-stamp-ext-hdr
• Discussion:
  • This draft mirrors the IPv6 extensions draft for MPLS data planes. Revisions aligned the draft with the IPv6 version (specifically on egress node behavior and header control sub-TLVs).
  • An open-source implementation is available. The authors requested working group adoption.

11. Residual Bit Error Rate (BER) using STAMP

• Presenter: Rakesh Gandhi
• Discussion:
  • Focuses on measuring bit errors escaping FEC/CRC using STAMP padding patterns. A new TLV was added to detect burst bit errors.
  • Both Cisco (experimental) and open-source implementations exist.
  • Feedback: Richard Foot and Lichang supported early IANA code point allocation and adoption. Lichang suggested setting up a GitHub repository to track issues.

12. Route Origin Validation (ROV) Benchmarking for Routers

• Presenter: Linhui Sun
• Discussion:
  • Designed to measure the latency, scalability, and robustness of routers performing ROV.
  • Revisions based on SIDOPS mailing list feedback expanded the VRP dataset characteristics, clarified update state changes, and refined BGP convergence test cases.

13. AI Fabric Benchmarking / Switching Efficiency Framework

• Presenter: Wei Chang Sun / Student
• Slides: AI Fabric Benchmarking Suite
• Discussion:
  • Proposed a mathematical "switching efficiency" framework (eta) to evaluate how effectively AI network switching resources translate to computational progress in LLM training.
  • Feedback: Karsten noted that the draft reads like an academic paper rather than an IETF standard and raised concerns about potential AI-generated text in Section 5. Wei Chang Sun clarified that the goal is to define a high-level industry metric (similar to Power over Ethernet) that hardware vendors and operators can use to optimize network fabrics, and they intend to adapt the concepts into practical testing tools.

14. Lightning Talks (Proposed Work)

• Packet Loss Monitoring Framework for List of Congestions: An unidentified presenter summarized a framework to detect and isolate packet loss events in network devices without modifying packet headers.
• Active Packet Loss Interpretation Considerations (Presented by Linhui Sun, referencing draft-liu-ippm-active-loss-considerations): Proposes separating observed active measurement losses from inferred root causes (e.g., congestion, routing changes, or security filtering) using on-path telemetry context.
• AI Agent Security Benchmark (Presented by Mailington, referencing draft-han-bmwg-agent-security-benchmark): Outlined a 4-dimension framework (Foundation Model, Interaction, Operation, and Basic Security) to evaluate security risks in AI agent deployments.
• AI Agents Benchmarking in Network Operations (Presented by Luis): Outlined a methodology to evaluate AI agents executing operational tasks (e.g., troubleshooting and configuration) based on metrics like completion time, success rate, and token efficiency. Boris Khasanov (in chat) highlighted that this work should coordinate with the newly proposed Agent Protocols BoF/WG.

Decisions and Action Items

• Decisions:
  • The WG will progress draft-ietf-ippm-alt-mark-deployment following the completed WGLC.
• Action Items:
  • Nalini Elkins: Update draft-ietf-ippm-encrypted-pdmv2 to explicitly request a new option type from IANA.
  • Chairs: Initiate early review requests (SEC, INT, OPS, TSVART) for draft-ietf-ippm-encrypted-pdmv2.
  • Thomas Graf: Coordinate with CCAMP chairs to check on any ITU-T G.7710 liaison requirements for the PM streaming drafts.

Next Steps

• Initiate a joint WGLC with OPSAWG for the Alternate Marking IPFIX extension draft.
• Continue discussions on the mailing list regarding the STAMP ECN alternative "CoS v2" proposal and its backward compatibility implications.
• Review the newly proposed joint IPPM/BMWG charter (BPM) on the mailing list.
• Issue a call for adoption on the mailing list for draft-ietf-ippm-stamp-ext-hdr and the STAMP Residual BER draft.

Related Documents

draft-fioccola-ippm-rfc7799bis, draft-han-bmwg-agent-security-benchmark, draft-ietf-ippm-alt-mark-deployment, draft-ietf-ippm-alt-mark-yang, draft-ietf-ippm-connectivity-monitoring, draft-ietf-ippm-encrypted-pdmv2, draft-ietf-ippm-on-path-active-measurements, draft-ietf-ippm-responsiveness, draft-ietf-ippm-stamp-cos-ecn, draft-ietf-ippm-stamp-ext-hdr, draft-liu-ippm-active-loss-considerations