1. Pages
2. …
3. Open Transport Configuration and Control Home
4. TAPI
5. TAPI Call-in Details and Notes

2023-04-18 TAPI Meeting notes

All

Preliminary agenda:

• Continue review Feature priority for TAPI "2.5"
• TAPI Hybrid - Equipment and OAM augments to TAPI 2.1.3
  • Feedbacks from TIP MUST
  • Photonic OAM vs. Inventory - RIA 1.2 (and 2.1) may make PM metrics as optional in Inventory UCs
• Path Computation UCs
• Model of the internal connectivity matrix of a ROADM
• OAM and Protobuf/gNMI encoding
• Continue the investigation on TAPI-IETF navigation
• Access to the photonic plug

Andrea Mazzini 

Andrea Mazzini shows the last pull requests to 2.4.1 and 2.1.x branches.

Liaison to MEF under preparation, regarding TR-548 v2.0, below one draft statement:

We are aware that MEF LSOC is specifying Streaming feature in the context of Service OAM and Service Function Testing for the Allegro, Interlude and Legato Interface Reference Points project. Given the interworking relationships between these IRPs and Presto, we would like to inform that ONF TAPI has recently delivered the version 2.0 of Streaming Reference Implementation Agreement (TR-548 v2.0).

• Nigel Davis we may also inform MEF about the Streaming evolution in 2.5
• Andrea Mazzini Streaming feature seems very coupled with the specific encodings/protocols.
  • Nigel Davis it is necessary the separation of concerns.

Esther strongly underlines the risks of having two (parallel and not compatible) development streams, for interoperability and possible confused message from TAPI team.

• Which is the decision process?
  • Andrea Mazzini answers that so far the consensus was always reached without the need of formal voting, in other words there were not strongly conflicting positions.
  • Nigel Davis our aim is to facilitate and enable, rather than decide.
• Esther announces that she will inform TIP MUST of the new TAPI planning, to get a shared position.
• TAPI team agrees that TIP MUST position will be a key driver of the TAPI delivery plan.
• Andrea Mazzini we are aware of the risks, nevertheless there are several implementations in version 2.1.3 which may benefit from small specific and backward compatible enhancements.
• Nigel Davis The 2.1.4 is intended as a side track, extremely not preferred. According to TIP MUST position, we may de-standardize the 2.1.4 delivery.
• Gabriele Galimberti Cisco has already made several extensions to 2.1.3, could we propose them for standardization? This also because 2.4.x implementations will not come any soon. Nigel Davis points out that Ciena has found the migration easier than expected.

Overview of the page Feature priority for TAPI "2.5"

• Path Computation, agreed that Nigel Davis and Andrea Mazzini will provide a starting point for discussion, leveraging existing UCs and previous contributions like

  • otcc2020.ND.010_TAPI-ConnectivityServiceComputationService.pptx

  • otcc2021.RC.001_TAPI-PathComputation.pptx

• Gabriele Galimberti we may consider to improve the SRLG format. For further analysis.

Andrea Mazzini shows an email by Roshan Joyce with some questions for clarification.

• After some discussion, it is agreed to remove the highlighted part of the following RIA statement, because it is adding more confusion than clarification:

TAPI models define Global objects and Local objects:

- A global object (an object that belongs to the GlobalClass) includes an uuid that is unique (not only in the scope of its containing parent/ancestors but also at least for the applicable TAPI context).

Andrea Mazzini notes that some CEP attributes may not be suitable for generic inventory, like PM Metrics.

• In fact, the PM Metric of a CEP may need specific operations which may slow down the inventory process, even when not really required for the specific operation context.
• Agreed that it is necessary to distinguish between different natures of the modeled items:
  1. Configuration parameters
  2. Stable resource state parameters (e.g. configured mapping types, thresholds, etc.)
  3. Not stable resource state parameters (e.g. operational state, PM Metrics, etc.)
• Each parameter type requires different access methods. See for example the ActiveCondition list of FaultManagementContext.
• Nigel Davis another example is the GET on OamJob instance, which may not always require also all contained current and history data instances.
• Nigel Davis another aspect to be considered is the urgency vs. scheduling needs.
  • Closed loop optimizations are different from post processed statistics.

Andrea Mazzini briefly shows the minutes of last week call (2023-02-14 TAPI Meeting notes)

Esther indicates a possible error in the "n" numbering of the presented example.

Gabriele Galimberti agrees. We will inform Ramon Casellas.