The meeting focused on orchestration strategies for multi-control plane environments, with Alan leading the discussion as the designated domain lead for this bi-weekly session. The team discussed supporting multiple control plane options including Kubernetes and VM-based solutions, with particular attention to OpenStack as a candidate for telco and broadband provider deployments. Key topics included identity and access management approaches, with debate around whether to implement a unified identity provider across hosting providers or allow each provider to manage its own identity system. The group also explored different operational models for tenant deployment, including managed Kubernetes control planes, namespace-based isolation, and VM-based approaches, though performance considerations and isolation requirements were noted as potential challenges for edge cloud environments. The discussion concluded with agreement to gather more requirements and define specific consumption models before making final implementation decisions.
- Alan: Start capturing the discussed orchestration models, IAM concepts, and operational models in a document for review in upcoming calls (target: before the next relevant orchestration discussion, potentially in two weeks).
- Alan: Draft a 2-3 page document outlining the proposed IAM/identity and access management approach, including basic roles and policies, for group review.
- Yoav/Team: Update the survey to include questions about expected levels of isolation (e.g., namespace vs. VM vs. bare metal) and gather feedback from relevant stakeholders.
- Alan: Describe pros and cons for each orchestration/consumption model (managed Kubernetes control plane, managed end-to-end clusters with namespacing, VMs, etc.) in preparation for the next working session.
- All: Communicate offline to coordinate whether Alan’s materials will be ready for the two-weeks-later call and adjust the agenda accordingly
- Antitrust Statement & Multi-Cloud Orchestration Strategy Discussion
- The team discussed orchestration strategies, focusing on supporting multiple control planes including Kubernetes and OpenStack.
- They explored the need for a wrapper API to handle provisioning across different hosting providers, with specific attention to compute availability, identity and access management, and metering/billing interfaces.
- The discussion included a comparison to public cloud architecture, proposing a tiered approach with hosting providers as regions, sub-regions as locations, and individual zones within each provider.
- Two identity management models were considered: either a common identity provider trusted by all hosting providers, or allowing each provider to maintain its own identity system.
Identity Management Framework Discussion
- The team discussed identity management approaches, with Yoav and Ben initially favoring promoting existing OAuth solutions while Alan proposed creating a standardized identity provider framework.
- Alan suggested developing a basic specification for common roles and policies across hosting providers, starting with coarse identity definitions that could be extended over time.
- The group agreed to move forward with creating a 2-3 page document outlining the proposed approach, with Alan taking the lead on describing the concept and establishing basic identity and access control frameworks.
Kubernetes Resource Consumption Options
- Alan discussed three options for consuming resources, focusing on the differences between managed Kubernetes in public cloud and other approaches.
- He explained the distinction between using shared resources, providing dedicated compute, and accessing existing clusters with high-rent namespaces.
- Alan emphasized that while VMs offer flexibility, they lack the intent-driven reconciliation and management benefits of Kubernetes.
Tenant Cluster Configuration Models
- Alan and Yoav discussed different models for tenant cluster configurations, including running Kubernetes on bare metal, VMs, or a combination of both.
- Alan expressed concerns about performance penalties when running on VMs and emphasized the importance of efficient resource utilization.
- Yoav shared their experience running Kubernetes on VMs, noting that despite best practices to mitigate performance impacts, customers still preferred strong isolation, leading to their implementation on VMs.
- Both agreed that the final approach should balance performance, isolation, and manageability requirements.
Kubernetes Deployment Models for Edge
- The team discussed different Kubernetes deployment models for edge cloud environments, focusing on isolation requirements and operational approaches.
- They identified three main scenarios: managed Kubernetes control plane with node allocation, namespace-based isolation, and VM-based deployment.
- The group agreed to gather requirements and define consumption models before making final decisions, particularly around identity access management and orchestration approaches.
- They noted that while public cloud models serve as a reference, edge environments may require different considerations due to limited resources and performance implications.
- The team decided to document pros and cons of each approach before determining whether to support multiple models or focus on specific options.
Next Meeting: Thursday, April 2, 2026 9am PDT