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

Compare with Current View Page History

« Previous Version 20 Next »

Topology Class Diagram

General discussion items

Approach

 This page is being used to explore concepts related to topology.

Where a concept is understood it needs to be put in the context of the OANP Information Model (Root). This may involve updates to the ONAP Resource IM. It is important that the topology model and the ONAP IM are consistent.

The topology model is inspired by TAPI. It will use classes, relationships and constructs from TAPI together with additions that make it what it needs to be.

 

This diagram shows how the topology model can be integrated with the ONAP IM through use of the Domain class. There was some discussion of this. The general approach is valid. Specifics and thinking are still required, for example, is there a need to a TopologicalService as a class distinct from a Service?

Top level construct

I am missing a class that can be used by services that need to include flow domains, connectivity and endpoints. I see it as useful and flexible to allow this.

For sure the TAPI model does not deal with services (customer and resource facing) – so I think that we should keep them loosely coupled via this missing class.

So just to play with a name here, lets call it a 'Reosurce'.


I'm not trying to describe the service model here, just show how it could relate to the resource side of the house, and why the 'Resource' is an elegant way to hide some of the complexity of what a service is really made up of.

The separation of the Service and Resource in the Topology reflects the current ONAP IM split. There is a Resource in the ONAP IM. It has a single sub-type, 'NetworkFunction'. Need to consider how to merge/align. For example a new TopologyEntity as a peer of the NetworkFunction with the FlowDomain, Connectivity and EndPoint as sub-types of the TopologyEntity...

Ownership

The parent page (Abstract Topology Model) describes a requirement for an association of a topology entity to the software that owns it for a given purpose (e.g.: LCM; Configuration)

This is also missing from the proposal. Again just the concept here – of course it needs to have relationships to FlowDomain and Connectivity. These can be managed from multiple perspectives (LCM; Configuration; Assurance; ...) by different products.


You could put the 'ManagementReference' as attribute(s) of the Resource, but I think it is (1) a type of resource in its own right and (2) potentially structured information required to interface with the actual manager instance.

The need for this is valid and useful. The 'managedBy', relationship may be promoted to the Resource level – to be less proscriptive. Note also that there will be multiple ManagementReference from a single Resource, potentially one for each management domain (e.g. Fault, Configuration, Performance subscription, etc. This also looks like a update to the 'Resource' in the ONAP IM.

Intent and services

Services

The separation of concerns (services and resources) is a principle that should be enshrined in the topology model.

By its nature a service is a simplified abstraction of what is used to realize it. The implementation details typically do not matter to a customer as long as the SLA is met. Bundling the service concepts into a resource model is going to make life more complex. Where there are service specific modeling needs, lets add them to the Service classes.

Makes sense, aligns with ONAP IM separation to date.

Agreed to remove the ServiceEndPoint and ServiceFlow/Intent from the sketch.

Intent

Intent is a wonderful new buzzword. It is only as good as the interpreter is. Eventually the intent needs to be translated into configuration of endpoints, connectivity and flow domains. Intent is also hierarchical, it will be evolved by layers of interpreters. Of course we need to keep the reference back to the intent for traceability (and possibly in scheduling). This can be achieved by having a simple hierarchical element that is referenced by the 'Resource'. So the participation of any resource in an 'intent' can be traced and interrogated regardless of the interpreter.


This is an alternative to the sketch, "ServiceFlow/Intent". More information is needed on what 'Intent' is. This part of the model will be parked for now.

Specific named classes

ManagedFunction

I think this is a type (specialization) of FlowDomain. Basically it is hierarchical in nature, i.e. value defined by an organization or standard that may be realized by virtual machines and or containers. It will contain termination points

Discussion started. Moved towards atomicity based on perspective. A VNF is atomic from a telecom CM perspective (you may create a rule on a firewall instance), but be decomposed into VNFC and subsequently container instances from a lifecycle or assurance perspective.




  • No labels