rtgwg

Summary

The RTGWG meeting covered updates on several drafts, including the Queueing YANG Model, SR network protection, and new scenarios for multi-cloud SD-WAN. Discussions focused on clarifying the scope and purpose of each draft, ensuring alignment with existing technologies, and addressing implementation complexities. Key topics included BFD session optimization, BGP extensions for multi-topology routing, and in-network computing.

Key Discussion Points

Queueing YANG Model: Updates included addressing YANG doctor comments, adding economics and definitions, language modifications, and a queue design diagram. Further enhancements were proposed to include a scheduling policy example and improve filter descriptions.
SR Network Protection: The presentation covered local and end-to-end protection mechanisms. The discussion focused on optimizing BFD session usage and clarifying the draft's scope as an informational document summarizing existing technologies. A suggestion was made to include a considerations section explaining why particular technologies were chosen for deployment.
Multi-Cloud SD-WAN: The presentation highlighted new scenarios requiring routing protocol enhancements for multi-cloud SD-WAN deployments. The discussion questioned the novelty of the presented concepts compared to existing SD-WAN solutions and emphasized the complexity of implementing BGP for multi-topology routing with the introduction of Beer. The need to demonstrate sufficient interest within the working group to justify the draft was highlighted.
Signaling in In-Network Computing: Two drafts were presented: the main specification and deployment considerations. The discussion clarified the relationship between this work and CATS, emphasizing that this work focuses on on-path computation between two endpoints where a function needs to occur before reaching the destination, while CATS focuses on selecting the computation point itself. Concerns were raised about the suitability of the proposed solution for collective operations due to issues like signaling and consistency. Suggestions were made to focus on the characteristics of the computation rather than the specific operations.

Decisions and Action Items

Queueing YANG Model: Prepare for working group last call.
SR Network Protection: Add a considerations section detailing technology choices and BFD optimization strategies.
Multi-Cloud SD-WAN: Engage the mailing list to gauge interest in the draft and refine the proposal based on feedback.

Next Steps

Continue discussions on the mailing list for all drafts.
Address comments and update drafts based on feedback.
Presentations on Friday.

Session Date/Time: 31 Mar 2023 00:30

rtgwg Meeting Minutes

Summary

The RTGWG meeting covered several draft proposals and updates, including alternative service routing, blockchain applications in BGP, a protocol assistant for troubleshooting, advanced topologies for data centers, tactical traffic engineering, and requirements for network fault detection. Several presentations were given, followed by Q&A and discussion.

Key Discussion Points

Alternative Service Routing (ROSA):
- Discussed the motivation for in-band, on-path service discovery as an alternative to DNS-based solutions.
- Presentation of updates to the draft, including new use cases (e.g., interconnect, mobile applications, video delivery), analysis of benefits, and relation to existing standards.
- Concerns about the destination header in IPv6 and potential solutions discussed.
- Debate about deploying ROSA over an IP network.
Blockchain for BGP:
- Explored potential applications of distributed consensus systems (DCS) like blockchain in BGP.
- Discussed key challenges: update latency, costs (communication), and working with inconsistent state.
- Potential use cases include avoiding origin AS assignments and validating updates.
- Feedback suggested the work may be better suited for the IRTF. Concerns were raised about applicability to existing authorities.
Protocol Assistant (PAS):
- Presented a protocol assistant for troubleshooting, a semi-centralized approach to gather troubleshooting data.
- Discussed use cases, including peer disconnect detection, route interruption detection, and OAM event correlation.
- Discussion of implementation as a new protocol vs. using existing GRASP mechanisms.
- Concerns about the query mechanism and notification of failures.
Dragonfly Flight Power:
- Discussion of Dragonfly and Dragonfly+ topologies for data centers.
- Focus on reducing network diameter, improving scalability, and enhancing path diversity.
- Concerns about path hunting and congestion control.
Tactical Traffic Engineering:
- Proposed a solution to detect congestion in real-time and move traffic using backup tunnels.
- Addressed recognition of congestion, flow manipulation, downstream congestion mitigation, and flow selection.
- Questioned action being local and not global.
- Concerns about out of order packets.
Requirements for Network Fault Detection:
- Presented a framework to accelerate fault detection for high-performance applications.
- Discussed the benefits of network-assisted fault detection.
- Questioned failure detection vs. notification.
GP6 Updates:
- Overview of the GP6 work and history.
- Valuable use case to justify the scope and value of API's.

Decisions and Action Items

Blockchain for BGP: Consider moving the work to the IRTF based on feedback received. Facilitate discussions with relevant stakeholders to assess feasibility.
All Presentations: The RTGWG requests future presentation to include problem statements that progress WG efforts.
All: Continue discussions on the mailing lists.

Next Steps

Authors to incorporate feedback into revised drafts.
Continued discussion on the mailing lists for each topic.
Set up virtual design team to work on GP6 API's.