Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

15 minsAdministrative

 TAPI Call: 2 hours

  • Review the updated TR-547-TAPI Reference Implementation Agreement_v1.1.docx
    • New use cases: 0d, 1g, 1h, 2a, 2b, 2c, 3d, 3e, 3f, 5d, 11a, 11b, 13b, 13c, 16a, 16b.
    • Finalize agreements on UC 1g/2b (server layer constraints).
      • Consider to suspend the discussion on constraints to proceed with the other new use cases.
  • Malcolm Betts will be reducing his level of participation in ONF TAPI.
15 mins

Review the updated TR-547-TAPI Reference Implementation Agreement_v1.1.docx

Agreed to specify the encoding of SAPI/DAPI in the RIA (UML type is string).

  • Ramon Casellas the string should be a 16 hex characters, lower case.
  • Agreed to add the "inter-domain-plug-id" attribute of ENNI-N NEP as name-value pair in 2.1.3 version, specifying the separator between SAPI and DAPI.
    • Delegate to the orchestrator the management of SAPI/DAPI combinations at the two ENNI terminations, e.g.:
      • East ENNI-N it is SAPI-East/DAPI-West
      • West ENNI-N is it SAPI-West/DAPI-East
120 mins

Review the updated TR-547-TAPI Reference Implementation Agreement_v1.1.docx

  • Use case 1g: Unconstrained PHOTONIC_MEDIA/OTSiMC over MC Service Provisioning

Andrea Mazzini shows the 2021-01-26 TAPI Additional Call - Meeting Notes, use case 1g.

  • Last week Ramon Casellas indicated that it could be possible to stretch the Restconf by specifying in the same POST more distinct roots.
    • In case, the POST response should include n x HTTP Header Locations.
    • Reaffirmed that Restconf prescribes that identifiers must be defined by the Client. It is not foreseen Server assignment of identifiers.
  • More general discussion on how the ConnectivityService creation is coded in the "Minimum subset required of TAPI RESTCONF Data API" table of RIA.
  • Eventually agreed that the current POST is correct with respect to Restconf specifications:
    • /tapi-common:context/tapi-connectivity:connectivity-context/
  • Considering REST, the above expression would indicate a modification of ALL the content of "connectivity-context", i.e. a complete replacement of the whole "connectivity-service" list. In Restconf, the above expression indicates an ADDITION of the specified node to the "connectivity-service" list.
    See https://tools.ietf.org/html/rfc8040#appendix-B.2, where "artist" is the new "connectivity-service" and "library" is the "connectivity-context":

B.2.1. Create New Data Resources

To create a new "artist" resource within the "library" resource, the client might send the following request:

POST /restconf/data/example-jukebox:jukebox/library HTTP/1.1
Host: example.com Content-Type: application/yang-data+json { "example-jukebox:artist" : [ { "name" : "Foo Fighters" } ] } If the resource is created, the server might respond as follows: HTTP/1.1 201 Created Date: Thu, 26 Jan 2017 20:56:30 GMT Server: example-server Location: https://example.com/restconf/data/\ example-jukebox:jukebox/library/artist=Foo%20Fighters Last-Modified: Thu, 26 Jan 2017 20:56:30 GMT ETag: "b3830f23a4c"
  • Then long discussion on the best way to provide server constraints of a ConnectivityService.
  • Andrea Mazzini presents a possible simplification of CSEP provisioning, where a new class, "ServerConstraint" is composed by CSEP:

  • In this way it is possible to recycle the technology specific augmentations of the generic CSEP without the need of additional CSEP instances, e.g.:

  • Karthik Sethuramanand Ramon Casellas highlight that the proposed approach is limited to the constraints expressed by CSEP augments. There are also e.g. routing and resilience constraints which may refer to server layer of the ConnectivityService under provisioning. 
  • Karthik Sethuramanand Ramon Casellas agree that the bottom-up provisioning is still preferable for complex multi-layer provisioning scenarios.
  • Andrea Mazzini recalls that regarding top-down provisioning, there is a requirement to create also server ConnectivityServices because:
    1. Alignment to bottom-up provisioning - if a given layer protocol foresees the ConnectivityService, then the ConnectivityService shall be instantiated regardless top-down/bottom up provisioning style.
    2. Editing capability of automatically created server layer connectivities.
  • Karthik Sethuraman points out that the best way to fulfil the above requirements is to provision all the ConnectivityServices, client and server one(s) together.
    • Using only "constraints" to drive the creation of server ConnectivityServices leads to two ways to edit a server ConnectivityService:
      • by constraints associated to client ConnectivityService, and
      • directly on the server ConnectivityService once it is created.
  • Andrea Mazzini post-meeting note: as already discussed on previous calls, it is not possible to provision a tree of ConnectivityServices, because the server ConnectivityService(s) may support other future client ConnectivityService(s). In other words, the tree structure does not work for ConnectivityServices.
  • Some conclusions:
    1. For Use Cases 1g and 2b check whether TAPI 2.1.3 can support them through simple amendments (or even just name-value pairs)
      • If not, i.e. the modifications to 2.1.3 appear complex, then remove these Use Cases from 2.1.3 scope.
    2. In general, for simple constraints it may be preferable to use mechanisms like the CSEP "ServerConstraint" presented above.
      • Post meeting note: Two-stage constraining of a server ConnectivityService:
        • first stage at provisioning time of client ConnectivityService (few details),
        • second stage at editing time of created server ConnectivityService (full details).
    3. Explore the provisioning of more ConnectivityServices in one operation, currently there are two possible solutions:
      1. combine more ConnectivityServices / "roots" in same Post,
      2. introduce the Order model.


  • Option a:



  • Option b:


Additional notes:

  • Allowing two distinct MCG CSEP instances may be a temporary solution for 2.1.3, but to be evaluated the forward compatibility.


  • Defining dedicated data structures for server layer constraints may be more forward compatible but to be evaluated the detailed impact on 2.1.3 version.


...