You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 14 Next »

 

Attendees:

AT&T - Brian Freeman

Orange - Olivier Augizeau, Ludovic Robert, Matthieu Geerebaert, Oliver Renais

Fujitsu - David Allabaugh, ravi rao, Raghavan Subramanian, Denise Provencher, Deepak Patel


Notes:

  1. Agenda
  2. Olivier A reviewed requirements for interconnect services based on discussion with the marketing and inter-carrier groups at Orange
    1. Reviewed three cases for cross-domain connections: (1) domain = vendor, and interconnection is required between vendor domains within a carrier; (2) domain = network managed by an operational unit with its own ONAP instance; and (3) domain = telco.
    2. Discussed whether #2 and #3 are effectively the same use case. For Orange, they are the same.
    3. Subsea cables are often managed by consortiums and access typically requires lengthy negotiation.
    4. Requirements include the need to monitor that the original service constraints are still being met (for instance, latency or sovereignty after a failure that causes a re-route)
    5. Need to link L0/L1 to the IP layer. Need a strategy for managing across layers, especially in failure cases. Can you dynamically create new L0 links (e.g. between routers) when there is an IP backbone failure, using an automated patch panel?
    6. General agreement that this is a complex problem and that we need to identify a small but valuable first step.
  3. Dave shared his understanding of MEF Interlude support for service provider interconnect. Interlude seems to assume that the BSS layer is handling selection of the interconnect point.
  4. Discussion of scope for this project:
    1. Part of what we need to do is model L0/L1 services in ONAP
    2. MEF interconnect model could be used as the basis of our work
    3. Consider swim lanes of effort:
      1. Modelling
      2. KPIs – what is needed, how should they be collected?
      3. Controller(s) for L0/L1 – functions required, APIs, where in the architecture?
  5. Next meeting –  , same time as May 9 meeting
  6. Action – all to review the business requirements slides for presentation to the use case subcommittee; be prepared to review on the 16th

 

Attendees:

AT&T - Martin Birk, Brian Freeman

Orange - Eric Debeau Olivier Augizeau , Olivier Renais

Fujitsu - ravi raoDenise Provencher  David Allabaugh Raghavan Subramanian , Dwayne Reeves, Deepak Patel


Notes:

  1. Raghavan shared the child page he created for project artifacts. He will add a child page for meeting notes (this page) and list the names of project members.
  2. Ravi shared his updated slides. There is general agreement to split the workflow into phases and separate onboarding from service design and instantiation.
    1. Onboarding
      1. “Onboarding” terminology slightly confusing when applied to topology discovery; you are actually onboarding the assets of the controllers.
      2. Agreement to split onboarding step into (1) controller onboarding (configure IP address, API to use, etc.), and (2) asset/topology discovery (infrastructure prep for service order)
      3. At design-time, artifacts required to define an external controller to discover pertinent parts of the domains (end-points, network models etc.)
      4. Current flows not very aligned with what ONAP uses (Orange concern about using CDS everywhere)
      5. Can we leverage anything from CCVPN discovery? CCVPN currently does not support discovery of a domain controller, but may have a useful model.
      6. SDN domain controllers support tens of thousands of assets so should we be treating/modelling it as a PNF? Does model break down sue to assumptions about scale of individual PNFs?
      7. We might want to set up infrastructure manually in the first release
    2. Service Design and Instantiation
      1. Service definition of optical using the same definition that ONAP information model
      2. The model defined at design-time is distributed by SDC via DMAP to SO, SDN-C, A&AI, Policy, CLAMP etc., each of which store locally for use in run-time
      3. How is the service request decomposed? Who decides little “a” and little “z” (the ENNI)? OOF?
      4. Should we use SNIRO emulator (running in Docker containers) for managing homing/allocation in service instantiation flows (OOF reference)?
      5. Will there be a serviceability check in an upper layer system, so that we know that big “Z” is outside of SP1’s domain?
      6. Inter-domain path computation SW piece is missing (which inter-connects area available & will we used)
      7. How will the service request from BSS systems look?
        1. will it be the decomposed service request b/w end-points within the one ONAP SP instance (A > a and z > Z for example) [Likely option]
        2. or will BSS send E2E service request & SO will identify & decompose end-points across SP partner domains
      8. We need to incorporate E2E business constraints into the service definitions - global availability, latency, geographical constraints, automatic restoration
      9. Scope of use-case is very simple right now, but more complex scenarios as probable. We can likely start off small.
      10. Need to check Interlude work – they may have addressed this
  3. Preparation for End User Advisory Group meeting in May
    1. Brian/Olivier/Eric will facilitate getting a spot on the agenda


 

Attendees:

AT&T - Martin Birk, Brian Freeman

Orange - Olivier Augizeau , Olivier Renais

Fujitsu - David Allabaugh, ravi rao, Raghavan Subramanian, Denise Provencher

Notes:

  • Scope of L1 interconnect use case does not include business negotiations (e.g. pricing).These functions should occur at a higher layer. However, we need to ensure that APIs support all business requirements. For example, price/cost may affect path computation.
  • Reviewed and commented on use case proposal & flow. Discussion regarding use of CDS; Brian noted that there is a learning curve.
  • Domains within a carrier: AT&T would probably use manual links between domains. Orange would have separate ONAP instances for each country plus international network; could use L1 interconnect between countries.
  • Agreement that use case supports a business need to automate L1 interconnects. Olivier A volunteered to document business requirements.
  • Need a place for artifacts. Suggestion to create a child page in wiki.onap.org.
  • Target Frankfurt release (since El Alto focused on stability). Need to get this on Alla Goldner’s use case list. Suggestion to present use case at User Advisory Group  at the end of May.


  • No labels