The Edge WG Call focused on updating a draft document, with Sanjay Mishra committing to finalize changes by the end of the week. The team discussed the implications of combining IPv6 and IPv4 into IPv11. Alan Arolovitch proposed a new protocol named after a musical group known for going to 11. The meeting also covered the orchestration system, emphasizing the need for a uniform API, federated identity, and role-based access. Alan outlined the scope of deliverables, including inventory discovery, provisioning, and service discovery APIs. The team debated the complexity of network capacity and latency metrics, ultimately agreeing to simplify the model for edge caching.
Action Items
- Draft API definitions for deliverables D1, D2, and D6 (authentication, role model, and service discovery/registry) once the overall scope is agreed, so the team can start implementing these APIs.
- Add diagrams (e.g., session flows) to the orchestration document before the next call to clarify the proposed architecture and flows.
- Vet the messaging proposal with a few colleagues to validate the design and ensure it is not overly complex or impractical before presenting it in a future call.
- Update the shared draft document (including adding the requested diagram) based on the changes discussed, aiming to complete the textual cleanup by the end of the week or over the weekend.
- Coordinate with the Edge WG to ensure the orchestration document is read and that participants leave comments on it before the next call.
Antitrust Statement and Group Discussion
- Glenn Deen explains the discussion about IPv6, IPv4, and the need to combine them into IPv11 due to availability.
Discussion on Orchestration System and Messaging
- Yoav Gressel from Qwilt introduces the plan to discuss the orchestration system and messaging
- Yoav clarifies that the messaging is a suggestion for the team to work on, which might combine into the orchestration system later.
- Alan Arolovitch reviews the document, focusing on the goals and principles, including the uniform interface and federated identity.
- Discussion on the need for discoverable orchestration and extensible service architecture.
- Alan suggests removing the requirement for discoverable orchestration as it might not be necessary.
- The conversation shifts to the scope of standardization, including authentication and role-based access within tenants.
Detailed Scope and Deliverables
- Alan outlines the scope, including standardizing authentication, role model definitions, inventory discovery API, and provisioning API.
- Discussion on the need for service discovery and register API to enable tenants to discover and use services.
- Alan emphasizes the importance of keeping boundaries clear and allowing incremental adoption.
- The conversation touches on the need for a simple and extensible model for request routing and orchestration.
Challenges with Network Capacity and Latency
- Glenn Deen raises concerns about the complexity of calculating network capacity and latency.
- Alan suggests keeping it simple by focusing on compute capacity and static egress limits.
- Yoav Gressel adds that the network operator should provide guidelines for high availability and other constraints.
- Glenn Deen proposes using hop count as a proxy for latency and capacity, emphasizing the importance of low latency for edge caching.
Modeling Request Routing and Orchestration
- Alan suggests building a model to represent the orchestration and request routing, including different tiers of compute capacity.
- Discussion on the need for a static inventory of resources and the dynamic mapping of requests to available resources.
- Glenn Deen emphasizes the importance of low latency and the need for a simplified model for edge caching.
- The conversation concludes with a plan to review and refine the document, including adding diagrams and session flows.
Next Call: May 7th, 2026 @ 9am PDT