6.3.13 NWDAF discovery and selection

Multiple instances of NWDAF may be deployed in a network.

The NF consumers shall utilize the NRF to discover NWDAF instance(s) unless NWDAF information is available by other means, e.g. locally configured on NF consumers. NF consumers may make an additional query to UDM, when supported, as detailed below. The NWDAF selection function in NF consumers selects an NWDAF instance based on the available NWDAF instances.

The NRF may return one or more candidate NWDAF instance(s) and each candidate NWDAF instance (based on its registered profile) supports the Analytics ID with a time that is less than or equal to the Supported Analytics Delay.

The following factors may be considered by the NF consumer for NWDAF selection:

• S-NSSAI.
• Analytics ID(s).
• Supported service(s), possibly with their associated Analytics IDs.
• NWDAF Serving Area information, i.e. list of TAIs, for which the NWDAF can provide services and collect data; for each item of this list, a weight may be defined in the NWDAF NF profile to indicate the priority of the NWDAF to cover the TA. If there exists AoI, then the NWDAF whose serving area covers the AoI may be selected.

NOTE 1: If all services provided by one NWDAF do not support the same Analytics ID, the NWDAF registers the Analytics IDs of the services at the service level.

NOTE 2: Analytics ID(s) at service level take precedence over Analytics ID(s) at NF level.

NOTE 3: For discovery of NWDAF supporting Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo services, the Analytics IDs at the NWDAF NF profile are used.

• (only when DCCF is hosted by NWDAF):
• NF type of the data source.
• NF Set ID of the data source.

NOTE 4: Can be used when the NWDAF determines that it needs to discover another NWDAF which is responsible for co-ordinating the collection of required data. The NWDAF does a new discovery for a target NWDAF via NRF using NF Set ID or NF type of the data source, or using Area of Interest.

NOTE 5: For discovery of NWDAF supporting Nnwdaf_DataManagement service, at least the NWDAF Serving Area information from the NWDAF profile are used.

NOTE 6: The presence of NF type of data source or NF set ID of the data source denotes that the NWDAF can collect data from such NF Sets or NF Types.

• Supported Analytics Delay of the requested Analytics ID(s) (see clause 6.2.6.2).
• Vendor ID(s) of potential target AnLF(s), e.g. used in transfer procedure.

In the case of multiple instances of NWDAFs deployment, following factors may also be considered:

• NWDAF Capabilities:
• Analytics aggregation capability.
• Analytics metadata provisioning capability.
• Accuracy checking capability (i.e. analytics accuracy checking capability for the AnLF and/or ML Model accuracy checking capability for the MTLF as defined in clause 5C.1 of TS 23.288 [86]).

Applicable when NF consumer cannot determine a suitable NWDAF instance based on NRF discovery response and when NWDAF registration in UDM is supported, as defined in clause 5.2 of TS 23.288 [86]: NF consumers may query UDM (Nudm_UECM_Get service operation) for determining the ID of the NWDAF serving the UE. The following factors may be considered by NF consumers to select an NWDAF instance already serving a UE for an Analytics ID:

• SUPI.
• Analytics ID(s).

When selecting an NWDAF for ML model provisioning, the following additional factors may be considered by the NWDAF:

• LMF-based AI/ML positioning indication (indicating that the NWDAF containing MTLF supports ML model training for LMF-based AI/ML positioning).
• The ML model Filter information parameters S-NSSAI(s) and Area(s) of Interest (see clause 5.2, TS 23.288 [86]) for the trained ML model(s) per Analytics ID(s) and ML Model Interoperability indicator per Analytics ID, if available.

When selecting an NWDAF that supports Horizontal Federated Learning (HFL), the following additional factors may be considered by the NWDAF:

• Time Period of Interest: time interval [start…end], during which the Federated Learning will be performed.
• when selecting HFL client NWDAF:
• FL capability type as HFL client NWDAF per Analytics ID.
• NF type(s) of the data source(s) where data can be collected as input for local model training.
• NF Set ID(s) of the data source(s) where data can be collected as input for local model training.
• ML Model Interoperability indicator.
• when selecting HFL server NWDAF:
• FL capability type as HFL server NWDAF per Analytics ID.
• The ML model Filter information parameters S-NSSAI(s) and Area(s) of Interest (see clause 5.2 of TS 23.288 [86]) for the trained ML model(s) per Analytics ID(s), if available.

When selecting an NWDAF that supports Vertical Federated Learning (VFL), the following additional factors may be considered by the NWDAF:

• Time Period of Interest: time interval [start…end], during which the Vertical Federated Learning will be performed.
• When selecting VFL client NWDAF:
• VFL capability type as VFL client NWDAF per Analytics ID.
• VFL Interoperability Indicator per Analytics ID.
• Optionally, supported Feature IDs per Analytics ID.
• Optionally NF set ID(s) of the data source(s).
• Optionally the Serving Area information.
• When selecting VFL server NWDAF:
• VFL capability type as VFL server NWDAF per Analytics ID.

When selecting a NWDAF for roaming case, the detailed mechanism is defined in clause 5.2 of TS 23.288 [86].