Session Date/Time: 19 Mar 2025 02:30

maprg

Summary

The MapRG session at the IETF in Bangkok featured presentations on various network measurement topics. Key discussions included RPKI collateral damage, Al4S congestion control deployment, Leo satellite network topology, ANICA flipping impacts on web browsing, and quality outcome metrics for network performance. The session also included a brief update on hackathon activities.

Key Discussion Points

RPKI Collateral Damage:
- Discussion on how networks deploying RPKI can still be vulnerable to hijacking if peering with non-RPKI-aware routers.
- Measurement study results indicating that a significant percentage of RPKI-protected networks are still victims of collateral damage.
- Proposed solution involving routers identifying non-RPKI-aware neighbors and avoiding forwarding traffic to them during hijacks.
Al4S Congestion Control:
- Analysis of Al4S performance in partially deployed environments with various bottleneck types and competing congestion control algorithms.
- Concerns raised about the coexistence of Al4S with BBR in dual-queue configurations.
- Recommendation to application developers to be cautious about deploying Al4S without understanding the network path and potential bottlenecks.
- Discussion on the importance of scalable congestion control on end hosts, particularly in asymmetric bandwidth scenarios.
Leo Satellite Network Topology:
- Exploration of the impact of various LEO satellite constellation parameters (altitude, number of satellites, inclination) on network performance (latency, stability).
- Findings indicated the existence of thresholds for satellite density and the importance of aligning orbit inclination with user endpoint geographic angles.
- Discussion on how terrestrial routing plays a role in practice, versus the simulation assumption of always bouncing back to the satellite.
ANICA Flipping:
- Re-evaluation of ANICA flipping and its impact on DNS root name servers and CDN performance.
- Measurement study showing that ANICA flipping can cause significant extra delay in web browsing.
- Discussion on potential causes of ANICA flipping (eCMP, BGP multi-path) and potential mitigation strategies.
- Suggestion to consider Quick and its ability to avoid packet or IP-tuple-based routing using connection IDs to help ANICA flipping.
Quality of Outcome:
- Presentation of the quality outcome metric for bridging the gap between network measurements and application performance.
- Simulation study on how sampling rate and accuracy affect the quality outcome score.
- Findings highlighting the importance of choosing the correct sampling parameters based on the use case and the type of network problems being detected.

Decisions and Action Items

MapRG and Happy Working Group Collaboration: MapRG will team up with the Happy Working Group and call for measurement results having to do with happy eyeballs implementations.
Hackathon Participation: Promote participation in hackathon events, where attendees can test and measure application support for IPV6 transition networks using the IPV6 test pod.
Engage in IPPM Working Group: Get more people engaged with this quality outcome document and provide feedback from application developers to see if it is useful.

Next Steps

Promote testing and measurement of Happy Eyeballs implementations.
Continue collaboration between MapRG and the Happy Working Group.
Continue to develop and refine the Quality Outcome metric within the IPPM working group.
Engage more application developers in testing and providing feedback on quality outcome.
Investigate further the specific implementation decisions of proxies related to the RFC requirements.