Topics

HRK: I'm showing my thought process work here. It's anything resembling a schedule by any stretch of the imagination, and the deep dives into issues and specific questions reflect only my own curiosity and what troubles me. This is merely a starting point, and I hope that the Programming Committee does what it will. We also need to build in time to get to some planning and outcomes.....

CNF Topics 

• Taking Anuket RA and RC activities into workloads
  • Is it sufficient to come up with a basic framework to demonstrate functionality on an example RI?
    • Necessary – is it sufficient?!
    • Tortured metaphor: we've got a good foundation, but need to address plumbing, HVAC, walls, and flooring
  • Where does performance come in?
  • How do we get into more challenging interop?
    • CNI
    • Other networking frameworks (eg., Istio, Envoy)?
    • Networking abstraction – how do we push this past the goal line, publish, and test?
    • Data Plane
      • If performance tuning is a source of interop issues, how do we address these?
      • Are data plane issues as opaque to workloads as claimed (should we do a survey, or testing experiment to have non-anecdotal data)?
    • If this is too much work, what is the highest priority activity that will reduce the cost due to re-work for end users, platform vendors, workload (CNF) vendors?
    • How do we avoid the mistakes of the past? Let's at least avoid those and create new ones.
• Given the previous topic, what is the highest priority testing a badging program and enforce and provide?
• CNF vendor input summit
  • How do we ask better questions to the CNF vendor community? 
  • Bring as many CNF providers as possible (perhaps with a bit of pre-prepared questions)
    • Vendors who have deployed within the 5GSPB context
    • Magma
    • O-RAN SC implementations and vendors
    • OAI
    • Workload BUs rather than platform BUs in our NEP members
    • Active CNF vendors (e.g. Matrixx, Affirmed, Metaswitch)
    • LF Edge application requirements
  • What guidelines, testing, information would be helpful for them such that they don't have to re-do for each platform/end user
  • Are we talking how to give guidelines and test for them against the platform requirements, or are there specific testing regimes for applications?
  • Assumption: CNF to specific platform 1:1 as a deployment requirement does not and cannot scale.

Architecture Topics

• On-boarding: Alignment across ONAP and EMCO, taking into account the role played by k8s.
  • Assumption for ingress: Helm v3 
  • To the extent it needs to be extended/constrained, both groups should define together, and we can determine if and where it belongs upstream.
  • How does this discussion feed into Anuket and Anuket Assured program? Infrastructure→Applications>Onboarding>MANO>Infrastructure (end-to-end)
• Post-onboarding – Day N – where and how are ongoing metrics collected and how to they feed into relevant parts of the stack....
  • What is considered in or out of the stack in various architectures....
• Conceptual flow diagrams
  • Where do the focus areas of both EMCO and ONAP land? What is shared? What is not? What is gray? Beyond marketing......
  • Before we go into powerpoint comparison, I would suggest choosing a discrete set of  applications (including at least RAN and a webapp....mostly b/c those could overlap in a hybrid environment, and then also each group chooses one application that is specifically optimized for them – e.g., Core, and perhaps something that gets close to IOT). Map out for each application how the flow works in ONAP only, EMCO only, and a choice or two in hybrid scenarios. From there we can start to look at how we want to talk to the market, figure out what technical works needs to be done, and how.