ONAP and ETSI Data Model Mapping

  ONAP ETSI-Alignment Modeling Hierarchy

• For OSS Service-level modeling, continue to use org.openecomp.resource.abstract.nodes.service
• For NS modeling, use SOL001 tosca.nodes.nfv.NS as SDC AID DM NS node type
• The OSS-Service model references/includes associated NSs (1:M)
• In ONAP, tosca.nodes.nfv.NS references, a new VNF node type, org.openecomp.resource.abstract.nodes.ETSI.VNF (1:M) and a new PNF node type, org.openecomp.resource.abstract.nodes.ETSI.PNF (1:M) are used
• In ONAP, tosca.nodes.nfv.NS includes SOL001 tosca.nodes.nfv.NsVirtualLink (1:M)

VNFD Mapping to SDC AID DM

SOL001 VNF mapping to/from VNF SDC AID DM

• ONAP previous analysis
  • There is no 1:1 mapping between the SOL001 tosca.nodes.nfv.VNF and org.openecomp.resource.abstract.nodes.VF.
    • SDC VF node types has very few properties.
    • a lot of common properties are added to each node type created by SDC
    • new common properties are mainly about networking and ONAP specific properties
  • There is no clear mapping logic between the current node types
    • SDC creates new data type as ONAP deployment configuration.
  • In ETSI, the descriptor_id is used to reference to the VNF, but in ONAP, the UUID is used to reference to the VNF.
• Solutions
  • map between SOL001 VNF and SDC AID DM VNF selectively.
    • data in the SOL001 that would be considered Cloud Provider automation data could be remained unmapped
    • Examples of Service Provider data that would need to be mapped in order to drive ONAP automation would include:
      • HPA requirements needed by OOF to determine cloud feasibility
      • Input parameters passed to the Cloud provider
    • ONAP wants to distinguish between the mode/data required to drive ONAP functionality from a Service Provider perspective from any ancillary mode/data required to drive Cloud Provider automation.
  • UUID and invariantUUID must not be modeled as metadata type, per TOSCA YAML because the data provided within metadata, maybe ignored by TOSCA orchestrator and should not affect runtime behavior. 
  • Identify Cloud Provider automation-related data and remove it from the mapping
  • Identify Service Provider data and map it to SDC AID DM
  • ONAP SO NFVO uses SOL001 VNF; other ONAP runtime components could use the new mapped org.openecomp.resource.abstract.nodes.ETSI.VNF

Current VNF Modeling in ETSI and SDC

Solutions:

There are two options. For now, we chose the option A. The option B is under discussion.

Option A (chosen):

Define a new data type based on the org.openecomp.resource.abstract.nodes.VF with ETSI SOL001 VNF data type attributes.

• Make the org.openecomp.resource.abstract.nodes.ETSI.VNF a superset of both tosca.nodes.nfv.VNF and org.openecomp.resource.abstract.node.VF
• During VNF onboarding, SDC copies SOL001 VNF attribute contents to the corresponding attributes in the org.openecomp.resource.abstract.nodes.ETSI.VNF
  • In Honolulu, SO NFVO, VFC and SVNFM get those SOL001 VNF attributes from the descriptor, not from AAI. So, AAI schema changes are not expected in Honolulu.
• SOL001 VNF attributes in SDC AID DM VNF will be visible to SDC UI, so SDC UI can change the attributes.
  • But the onboarded vendor ETSI package will note be changed by the SDC UI users in Honolulu.
  • Since SO NFVO, VFC and SVNFM use only the original ETSI package, those changes will not be used in Honolulu;
  • For the Honolulu release, it is under consideration
    • to sync up between those modified SOL001 VNF attributes and the vendor ETSI Package attributes 
    • to reflect those modified SOL001 VNF attributes in the orchestration
• ONAP specific attributes that are inherited from the org.openecomp.resource.abstract.nodes.VF will be filled up by SDC (design time)
  • Those attribute contents will not be mapped back into the SOL001 VNF (reverse mapping). For that case, only the SOL001 VNF corresponding attributes will be copied

SOL001 VDU mapping to/from VNF SDC AID DM VFC

Current SOL001 VDU vs. SDC AID DM VFC

• no 1:1 mapping
• note: the SDC AID DM VFC represents design-time VFC (like VDU), not VFC instances in runtime

Solutions for VDU and VFC mapping

• Make the org.openecomp.resource.abstract.nodes.ETSI.VFC a superset of both tosca.nodes.nfv.Vdu and org.openecomp.resource.abstract.nodes.VFC
• Note: the org.openecomp.resource.abstract.nodes.VFC represents design-time VFC, not VFC instances
• During VNF onboarding, SDC copies SOL001 VDU attribute contents to the corresponding attributes in the org.openecomp.resource.abstract.nodes.ETSI.VFC
  • In Honolulu, SO NFVO, VFC and SVNFM get those SOL001 VDU attributes from the descriptor, not from AAI. So, AAI schema changes are not expected in Honolulu.
• SOL001 VDU attributes in SDC AID DM VNF will be visible to SDC UI, so SDC UI can change the attributes.
  • But the onboarded vendor ETSI package will note be changed by the SDC UI users in Honolulu.
  • Since SO NFVO, VFC and SVNFM use only the original ETSI package, those changes will not be used in Honolulu;
  • For the Honolulu release, it is under consideration
    • to sync up between those modified SOL001 VDU attributes and the vendor ETSI Package attributes 
    • to reflect those modified SOL001 VDU / VFC attributes in the orchestration

New SDC AID DM VFC type (org.openecomp.resource.abstract.nodes.ETSI.VFC)

SOL001 3.3.1 VNFD Mapping from/to SDC AID DM VNFD

SOL001 3.3.1 VNFD Template

tosca_definitions_version: tosca_simple_yaml_1_3

description:

imports:

data_types:

node_types:

topology_template:

    inputs: 

    node_templates: 

        VNF

substitution_mappings:

    capabilities:

    requirements:

SDC VFD Template

tosca_definitions_version: tosca_simple_profile_for_ecomp_1_0

descriptions:

imports:

data_types:

node_types:

topology_template:

    inputs:

    node_templates:

         vfc:

            type: org.openecomp.resource.vfc

        extCP:

            type: org.openecomp.resources.cp

    groups:

        VFModule_Base:

            type: org.openecomp.groups.VfModule

        VFModule_Expansion:

            type: org.openecomp.groups.VfModule

    workflows:

    policies:

        anti_collated_az_policy:

substitution_mappings:

    node_type: org.openecomp.resources.vf.<vf_name>

    capabilities:

    requirements:

SOL001 VNFD mapping to/from SDC AID DM VFD

VF-Module Initial Input 

The following is a summary of initial input from Gil Bullard (AT&T).

• Mapping Scope Challenges and Suggestions 

  • ONAP wants to distinguish between the mode/data required to drive ONAP functionality from a Service Provider perspective from any ancillary mode/data required to drive Cloud Provider automation.

    • See the ONAP MultiVim Proposal on the separation of concerns.
  • SDC AID  was created before a separation of concerns was envisioned, thus there are data structures in the SDC AID that would be considered Cloud Provider automation data and that we would likely take out of the SDC AID.
    • Examples of Service Provider data that would need to be mapped in order to drive ONAP automation would include:
      • HPA requirements needed by OOF to determine cloud feasibility
      • Input parameters passed to the Cloud provider
    • There is probably some data in the SOL001 that would be considered Cloud Provider automation data. Any such data could remain unmapped
    • SOL001 content needed to drive ONAP automation (SO, OOF, SDNC, AAI, DCAE, etc.) would need to be mapped
  • The VNFM should not be aware of the VF Module structure, though the current SO-SDNC interactions to get assignment are by VF Module, and the VF Module entity plays prominent in AAI. That means there needs to be two adaptation points as we have discussed:
    • An SDC function to extract VF Module information from the SOL001 VNFD prior to distribution to runtime
    • A VNFM Adapter function to flatten the VF Module structure prior to handoff to the VNFM

• VF-Module Deduction from SOL001

• • There is an assumption that SDC transforms the vendor provided VNF package into ONAP-compliant one; i.e., deducing VF Modules based on VNFD ScalingAspects and Delta.
  • If SDC supports the transformation in Dublin time-frame, the transformed CSAR will be imported to SO, and SO VNF- and VF-Module-level workflows will manage VNF and VF Module topology towards SDNC with the following changes - Input from Gil Bullard (AT&T)
    • Today the VNF-level workflow has an embedded per-VF Module loop that a) retrieves the SDNC assignments for that VF Module, and then b) sends those VF Module-level assignments down to the VIM (e.g., OpenStack); the loop then moves to repeat "a" with the next VF Module. 
    • The new VNF-level flow will have the following sequences:
      1. an embedded per-VF Module loop that only retrieves the SDNC assignments for each VF Module; because the VIM is hidden from SO's sight, beneath the VNFM Adapter/VNFM.
      2. After finishing the loop, the SO workflows will send a structure  to the VNFM Adapter that includes the aggregate assignments at the VNF level.
      3. The VNFM Adapter aggregates all the VF-Module level assignments and transforms the assignment data into SOL003 API parameters before sending them to SVNFM
         1. The VNFM Adapter would need to be able to parse the VNF-level assignments structure received from SO to obtain the per-VDUconnection point assignments information and any per-VDU parameter information (e.g., hostnames)
         2. In doing so, the VNFM Adapter would need to know to ignore the VF Module groupings of these assignments
         3. Further know how to map the ONAP data structure and parameter names into the ETSI (e.g., VM=VDU, VNFC=VNFC, vNIC=vNIC, etc.). Note that the above assumes that in ONAP, as in ETSI, there will be a one-to-one correspondence between VM/VDU and VNFC.

• 
  

  • Assumptions for deducing VF-Module from SOL001 (Gil Bullard's input)

• • • SOL001 concept of Aspect+ScalingDelta combination is one to one with the ONAP concept of VF Module.
    • SOL001 concept of VDU is one to one with the ONAP concept of A&AI vServer
    • SOL001 concept of a connection point associated with a VDU corresponds to the ONAP concept of the same name, as does the understanding of the meaning of “internal” versus “external” connection point.
    • ONAP-compliant SOL001 VNF Vendors will be obliged to name their HEAT files using a naming convention that encodes the SOL001 Aspect+ScalingDelta names
    • ONAP-compliant SOL001 VNF Vendors will be obliged to name their SOL001 Aspect+ScalingDelta parameters using a naming convention that readily maps to the corresponding HEAT properties.  
    • In addition, if AT&T has already deployed such a vendor’s VNF into its network, the HEAT naming will remain invariant, and (at least) the (AT&T version of that) SOL001 be written to match it.
  • What to do
    • ONAP will be extended to incorporate the constructs of Aspect and Scaling Level.  This includes SDC’s, SOs, and A&AI’s modeling of these constructs and A&AI's ability to capture and SO’s ability to set/update the "current scaling level" of a VNF for a given Aspect. 
    • If ETSI in their SOL001 VNFD had defined a "ScalingDelta" in a straightforward manner, i.e., in terms of the VNFCs that comprise it, then it would be very easy to extract VF Module information from the VNFD.  (I would like to see ETSI define "ScalingDelta" in this manner, as opposed to the current way they do so. )  However, given that they did not, ONAP SDC would need to be extended to derive the VF Module “structure” from the SOL001 document through the algorithm below.  “Structure” in this case includes the VDU topology, connection points and associated parameters.  This algorithm will:
    • Determine the set of Aspects and corresponding VDUs and associated ScalingDeltas (step_deltas) from the SOL001.
    • Determine the set of ScalingLevels associated with each Aspect and the ScalingDeltas associated with each.
    • Translate the VDU-centric representation of ScalingDeltas (step_deltas) as per SOL001 to come up with a ScalingDelta-centric representation that captures the number and type of VDUs associated with that ScalingDelta across the various VDU types.
    • Create a VF Module object that corresponds to each ScalingDelta-centric representation of a ScalingDelta calculated above.
    • Fill in the details of the VF Module object based on the SOL001 data to represent the VDU connection points, associated Networks (internal or external), and associated Parameters.
    • Determine if there is an artifact in the SOL004 package that is a HOT YAML whose file name corresponds (through a VNF vendor obligatory naming convention) to the Aspect+ScalingDelta from which this VF Module object was derived.  If so, associate that HOT file with the VF Module.
    • Name the VF Module based on a naming convention to capture the Aspect+ScalingDelta names
    • Determine and capture the mapping from each Aspect + ScalingLevel model for the VNF to the corresponding VF Module.
    • Given a desired state Aspect+ScalingLevel, will be able to calculate (from the SDC distributed mapping of Aspect+ScalingLevel to VF Module along with the current Scaling Level for this Aspect as per A&AI) the (ordered set of) VF Module(s) needed to take that VNF from the “current scaling level” to the desired level for that Aspect.
    • Note:  As an aside, SDC enhancements are being discussed. It is not clear if the SDC changes will be available in the Dublin time frame. some “stubbing off” SDC with a simulator could be suggested to at least prove in the run-time aspects of the solution.

Solution:

• SDC deduces the VF-Module from the SOL001 VNFD Policies>scaling_aspects>properties>aspects
• Additional VF-Module attributes are deduced as the following table
• SOL003 Adapter may need to transform the VF-Module back to the SOL001 VNFD policies for the scaling and healing requests from VNFM(s) 

org.openecomp.group.VfModule Attribute and SOL001 VNF Policies deduction