Markdown Version | Session Recording
Session Date/Time: 07 Nov 2024 17:30
rift
Summary
The RIFT working group meeting covered several key topics, including updates on milestones, discussions on Auto-ISIS, SRv6 extensions for RIFT, and a refresh on multicast in RIFT. A new charter has been approved, and the group is targeting several publications in the near future. Significant discussion revolved around the motivation and architecture of proposed extensions, particularly concerning SRv6 and multicast.
Key Discussion Points
-
Auto-ISIS (formerly Auto Flood Reflection):
- The document has been restructured to be more generic, applicable to L1, L2, or mixed ISIS topologies.
- The goal is to allow plug-and-play functionality for ISIS in disaggregated chassis environments.
- Working group asked for adoption.
-
SRv6 Extension for RIFT:
- Discussion centered around the need for traffic engineering in data center fabrics with ample bandwidth.
- Egress peer engineering was identified as a potential valid use case.
- Concerns were raised about the proposed key identifier mechanism potentially leading to tie-breaking issues.
- Suggestion to use a new tie type instead of stuffing SRv6 locator into the Prefix TIE.
- It was recommended the authors present SRV6 to spring working group.
-
Multicast in RIFT:
- The working group aims to resurrect work on multicast using bidirectional trees.
- The complexities and potential problems with broken connections at the tough and sub-tough levels were identified as reasons for previous stagnation.
- A new approach was suggested that leverages TOF information for a distributed algorithm in reverse flood reduction.
Decisions and Action Items
- Auto-ISIS: Working group asked for adoption of the Auto-ISIS draft, with minor tweaks.
- SRv6:
- Authors to incorporate a use case (e.g., egress peer engineering) to justify the need for SRv6 in RIFT.
- Authors to revisit the key identifier implementation to avoid tie-breaking problems.
- Authors to use a new tie type for SRV6 locator.
- Authors should look into Yang model for this extension
- Authors to present SRV6 solution to Spring working group.
- Multicast: Authors to consider the reverse flood reduction algorithm proposed by Tony P. and explore its feasibility.
Next Steps
- Authors of the SRv6 draft will revise the document based on the feedback received, including providing clear motivation and addressing technical concerns related to key identifiers and tie types. They will also create a young model for the extension
- Authors of the multicast draft will revisit their approach based on the "flood reduction in reverse" suggestion.
- Progress on the SRv6 extension should be presented at the next IETF meeting, at Spring working group and discuss progress.