Ike A Common MDA for 5GS Telecom and O-RAN Alliance SMO Rev PA06 Feb 2025.pdf

5G Advanced and B5G Networks
Management Data Analytics Function (MDA/MDAF) and
Service (MDAS MnS)
in
Telecom (4G/5G NSA) System
in
Telecom AI ML 5G Advanced specified
and
O-RAN Alliance SMO (Service Management and
Orchestration) Recommended SBA Architecture
Ike Alisson
2025 - 03- 05 Rev PA06

Table ofContents (ToC)
1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G NSA running on 4G LTE EPS CN and 5G SA)
2. O-RAN Alliance SMO Recommended SBA Architecture with conformance to 3GPP MDA
3. 5G MDA (MnS Producer and Consumer as specified for 4G LTE EPS and 5G) appliedfor5G AI ML
Annexes
1.5G AIML Model Training, Transfer, Split on Network Endpoints
2.3GPP 5G System (5GS) CN AIML enhancement with introduction of NF ReLF (Recommended Logical Function)
3.5G AIMLE (AIML Enablement) Architecture for AI ML Services
4.3GPP 5G System E2E KPIs
5.3GPP AI ML Consistency Alignment CN, RAN, AEF across all 3GPP WGs with Common AIML Terminology
6.5G specified MOCN with share RAN "direct & indirect connection" evolvement
7.5G User Identities evolvement
2

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G NSA running on 4G LTE EPS CN and 5G SA)

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G
NSA running on 4G LTE EPS CN and 5G SA)

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early
5G NSA running on 4G LTE EPS CN and 5G SA)

Figure: 5G CN NG-RAN Bearer Services QoS
Architecture
=>
1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G NSA running on 4G LTE EPS CN and 5G SA)

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G NSA running on 4G LTE EPS CN and 5G SA)

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early
5G NSA running on 4G LTE EPS CN and 5G SA)

1.Telecom MDA (with MnS) in 4G LTE EPS and 5G Systems (early 5G NSA running on 4G LTE EPS CN and 5G SA)

2. O-RAN Alliance SMO Recommended SBA Architecture with conformance to 3GPP MDA

3. 5G Management Data Analytics (MDA MnS Producer &Consumer as specified for 4G LTE EPS and 5G) applied for 5G AI ML

1. 3GPP all WGs co-ordination for cohesive & integrated Functional Operation for AI ML use in 5G ADVANCED RELEASEs at SYSTEM LEVEL

4. 3GPP 5G AIML related activities across all 3GPP WGs - 1

4. 3GPP 5G AIML related activities across all 3GPP WGs - 2 Common AIML Terminology

3. 5G Management Data Analytics (MDA MnS Producer &Consumer as specified for 4G LTE EPS and 5G) applied for 5G AI ML

96
The 5GS specifies the Generic Network Resource Information that can
be communicated between an MnS "Producer" and MnS "Consumer "for
Telecommunication Network Management purposes, including
Management of Converged Networks and Networks that include
Virtualized Network Functions (VNFs).
The 5GS specifies the Semantics of information Object Class (IOC)
attributes and relations visible across the Reference Point in a Protocol
and Technology neutral way. It does not define their Syntax and
Encoding.
The 5GS supports the Federated Network Information Model (FNIM)
concept in that the relevant Information Object Class (IOC) is defined in
this specification are directly or indirectly inherited from those specified
in the Umbrella Information Model (UIM) of Fixed Mobile Convergence.
Note. The presented NRM is applicable to Deployment scenarios using
the Service Based Management Architecture (SBMA) as defined in 5G
System Management and Orchestration Architecture Framework.
The Figures show the containment/naming hierarchy and the
associations of the Classes defined that are combined with the Figure
showing the Umbrella Information Model (UIM) Class Diagram.
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management

97
The 5G System specifies the Information Model and Solution Set for the
Network Resource Model (NRM) definitions of 5G NR, NG-RAN, 5G Core
Network (5G CN) and Network Slice (SST), to support the 5G System
Management for:
- Variety of 5G Radio Access Network (RAN) Functions and Features,
covering 5G Management for 5G NR Connectivity Options defined in
5G System for 3G, 4G LTE E-UTRA and 5G NR Multi-RAT Connectivity
and NG-RAN Architectural Options as specified by 3GPP.
- Variety of 5G Core Network (CN) Functions and Features defined by
3GPP in 5G System Architecture specification.
- 5G Network Slice (SST) and Network Slice Subnet.
The 5GS NRM Information Model defines the Semantics and Behaviour of
Information Object Class (IOC) Attributes and Relations visible on the
Management Interfaces in a Protocol and Technology Neutral way.
The 5GS NRM Solution Set defines one (1) or more Solution Set(s) with
specific Protocol(s) according to the Information Model definitions.
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management

98
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management

99
Directory: yang-models
Files:
_3gpp-5gc-nrm-affunction.yang
_3gpp-5gc-nrm-amffunction.yang
_3gpp-5gc-nrm-amfregion.yang
_3gpp-5gc-nrm-amfset.yang
_3gpp-5gc-nrm-ausffunction.yang
_3gpp-5gc-nrm-configurable5qiset.yang
_3gpp-5gc-nrm-dnfunction.yang
_3gpp-5gc-nrm-dynamic5qiset.yang
_3gpp-5gc-nrm-ep.yang
_3gpp-5gc-nrm-externalnrffunction.yang
_3gpp-5gc-nrm-externalnssffunction.yang
_3gpp-5gc-nrm-externalseppfunction.yang
_3gpp-5gc-nrm-FiveQiDscpMappingSet.yang
_3gpp-5gc-nrm-GtpUPathQoSMonitoringControl.yang
_3gpp-5gc-nrm-lmffunction.yang
_3gpp-5gc-nrm-n3iwffunction.yang
_3gpp-5gc-nrm-neffunction.yang
_3gpp-5gc-nrm-nfprofile.yang
_3gpp-5gc-nrm-nfservice.yang
_3gpp-5gc-nrm-ngeirfunction.yang
_3gpp-5gc-nrm-nrffunction.yang
_3gpp-5gc-nrm-nssffunction.yang
_3gpp-5gc-nrm-nwdaffunction.yang
_3gpp-5gc-nrm-pcffunction.yang
_3gpp-5gc-nrm-predefinedpccruleset.yang
_3gpp-5gc-nrm-QFQoSMonitoringControl.yang
_3gpp-5gc-nrm-scpfunction.yang
_3gpp-5gc-nrm-seppfunction.yang
_3gpp-5gc-nrm-smffunction.yang
_3gpp-5gc-nrm-smsffunction.yang
_3gpp-5gc-nrm-udmfunction.yang
_3gpp-5gc-nrm-udrfunction.yang
_3gpp-5gc-nrm-udsffunction.yang
_3gpp-5gc-nrm-upffunction.yang
_3gpp-5g-common-yang-types.yang
If the Class Managed Element and the underlying hierarchy is contained undera Sub Network all YANG
Modules containing IOCs that can be contained under the Managed Element directly or under other IOCs
contained by the Managed Elementand the YANG Module for Managed Element itself shall be mountedat
the mountpoint "children-of-Sub Network" in the YANG Module _3gpp-common-subnetwork.
IETF YANG Model describes the Mechanism that adds the schema treesdefined by a set of YANG
Modules onto a mount point defined in the schema treein another YANG module.
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management
5G Core Network (CN) YANG Definitions APIs specified and available at 3GPP

10
0
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management
5G CN NRM - Inheritance
The 5GS CN inheritance relationships that exist between IOCs, as depicted in the Figure below shows the Inheritance Hierarchy from IOC (Information Object
Class) Managed Function related to the 5G CN NF NRM. The Figure - 2 shows the Inheritance Hierarchy from IOC EP_RP related to 5G CN NF NRM.

10
1
The 5GS specifies the Generic Network Resource Information that can
be communicated between an MnS "Producer" and MnS "Consumer "for
Telecommunication Network Management purposes, including
Management of Converged Networks and Networks that include
Virtualized Network Functions (VNFs).
The 5GS specifies the Semantics of information Object Class (IOC)
attributes and relations visible across the Reference Point in a Protocol
and Technology neutral way. It does not define their Syntax and
Encoding.
The 5GS supports the Federated Network Information Model (FNIM)
concept in that the relevant Information Object Class (IOC) is defined in
this specification are directly or indirectly inherited from those specified
in the Umbrella Information Model (UIM) of Fixed Mobile Convergence.
Note. The presented NRM is applicable to Deployment scenarios using
the Service Based Management Architecture (SBMA) as defined in 5G
System Management and Orchestration Architecture Framework.
The Figures show the containment/naming hierarchy and the
associations of the Classes defined that are combined with the Figure
showing the Umbrella Information Model (UIM) Class Diagram.
2. 5G System Core Network (CN) and RAN Data Network Resource Model (NRM) for 5G System Management

10
2
The 5G System defined Class diagram of 5G Core Network (CN) Network
Functions (NFs) and defined the set of Classes (e.g. IOCs-Information
Object Class) that encapsulates the Information relevant for 5G CN NFs
NRM definitions.
The Relationships of relevant Classes is defined in UML. Subsequent
clauses provide more detailed specification of various aspects of these
Classes.
The Figure shows the 5G CN NF NRM Containment/Naming Relationship.

2. 5G CN Network Resource Model (NRM) for 5G System Management

3. 5G Management Data Analytics (MDA MnS Producer &Consumer as specified for 4G LTE EPS and 5G) applied for 5G AI ML

3GPP introduced Network Data Analytics Function (NWDAF) to support
-Network Data Analytics Services in 5G Core Network (5G CN),
-Management Data Analytics Service (MDAS) to provide Data Analytics at
the OAM, and
-Application Data Analytic Service (ADAES).
5G System provides support for AI/ML Services through
- AI/ML enabled support in NWDAF (through AnLF & MTLF)
- assisting the ASP/3
rd
Party AI/ML Application Service Providers (ASPs) for
AI/ML Model Distribution, Model Transfer, and Model Training for various UCs
and Applications (e.g., Video/Speech Recognition, Robot Control, Automotive,
AR/VR/XR, Immersive Applications, etc.).
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

3GPP System Architecture WG started from 3GPP 5G Rel-16 to investigate
the AI/ML support in the 5G System focusing on the main two (2) aspects.
1) AI-ML support in NWDAF
Network Analytics and AI/ML is deployed in the 5G Core Network (CN) via the
introducing of NWDAF consider the support of various Analytics types that can
be distinguished using different Analytics IDs, e.g., “UE Mobility”, “NF Load”.

Each NWDAF may support one (1) or more Analytics IDs and may have the
role of:
(i)AI/ML Inference called NWDAF Analytics Logical Function (AnLF), or
(ii)AI/ML Training called NWDAF Model Training Logical Function (MTLF) or
(iii)both.
For Federated Learning (FL) Use Cases for 5G, in Release 18, 3GPP defined
Federated Learning (FL) amongst different NWDAF MTLFs, where ML Model Training is running in Multiple Local MTLFs.
2) 5G Network assistance for AI-ML Services
5G CN AI/ML System enhancements are specified (in 5GS Architecture) for assisting the AI/ML Operations in the Application Layer (between one (1)
or more AI/ML Users and AI/ML Server).
5G CN NEF assist the AI/ML Application Server in scheduling available UE(s) to participate in the AI/ML operation (e.g. Federated Learning (FL)).
5G CN assist in the selection of UEs to serve as FL Clients, by providing a list of target member UE(s), then subscribing to the NEF to be notified about
the subset list of UE(s) (i.e. list of candidate UE(s)) that fulfil certain filtering criteria.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

In the Figure, it is depicted the Procedure about Consumers (NWDAF containing MTLF and NWDAF containing AnLF) to retrieve ML Models from an 5G
Analytics Data Repository Function (ADRF).
5G NWDAF ML Model retrieval from ADRF with NWDAF Federated Learning (FL) among Multiple NWDAFs
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF with an AnLF Analytics Accuracy Monitoring Procedures
In 5GS, the Analytics Accuracy Information comprises a set of parameters as defined in 5GS NDAF
Architecture and Flow Procedures specification.
When multiple5G CN NWDAFs are deployed, some NWDAFs could be specialized with the Analytics
accuracy checking Capability. When a 5G CN NWDAF containing AnLF has the Analytics Accuracy Checking
Capability, such a NWDAF is able to:
-Receive a Subscription or a request for Analytics IDs via Nnwdaf_AnalyticsSubscription_Subscribe or
Nnwdaf_AnalyticsInfo_Request Service Operation with the indication for activating the Mechanisms for
checking the Accuracy of such Analytics ID as defined in this specification
-Provide the Accuracy information to the Consumer via Nnwdaf_AnalyticsSubscription_Notify or
Nnwdaf_AnalyticsInfo_Request response service operation.
NOTE1: In this version of the specification, NWDAF containing AnLF can provide accuracy information to an NF consumer
that subscribes or requests for the analytics.
NOTE2: When receiving a subscription from an NF consumer that includes a request for accuracy information, the analytics
output and the accuracy information can be provided by NWDAF containing AnLF in a single notification or via
separate notifications.
NOTE3: When receiving a request from an NF consumer that includes a request for accuracy information, the analytics and
the accuracy information can be provided by NWDAF containing AnLF within the single response.
NOTE4: In this version of the specification, the subscription or request for accuracy information independently from
requesting an analytics output is not supported.
Depending on the specified triggers, the NWDAF containing AnLF starts the accuracy monitoring and
generation of Analytics Accuracy Information for an Analytics ID.
The Analytics Accuracy Information may be requested per Analytics ID and scoped using the same
parameters as those defined in the Target of Analytics Reporting as defined in clause6.1.3 and Analytics
Filter Information (e.g. for a specific area, specific slice) of the requested Analytics ID.
When the analytics accuracy checking is activated in an NWDAF containing the AnLF, the NWDAF could
store for a period of time the necessary information to determine the analytics accuracy and provide the
accuracy information to consumers when requested or use it for its internal processes.
The NW5G CN NWDAF P rocedures for Analytics Accuracy Information Subscription is used by 5G NF
"Consumers" of Analytics ID to subscribe to receive Analytics output and Analytics Accuracy Information
related to the requested Analytics ID for NF Consumer as shown in the Figure.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF with MTLF-based ML Model Accuracy Monitoring
procedure is where an NWDAF containing MTLF determines ML
Model degradation based on newly collected Test Data and retrain
or reprovision the existing ML Model.
The Figure illustrates the Procedure for Monitoring the Accuracy of
the provisioned ML Model using newly collected Data.
5G CN NWDAF containing AnLF provides Inference Data to
NWDAF containing MTLF for Model Accuracy Monitoring and the
NWDAF containing MTLF determines retraining or re-provisioning
of the ML Model.
5G CN NWDAF with MTLF-based ML Model Accuracy Monitoring
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

The 5G CN NWDAF can provide Observed Service Experience (i.e. average of observed Service MoS and/or Variance of Observed Service MoS indicating Dervice MOS distribution for Services such as Audio-
Visual Streaming as well as Services that are not Audio-Visual Streaming such as V2X and Web Browsing Services) Analytics, in the form of Statistics or Predictions, to a Service "Consumer".
The 5G CN NWDAF Observed Service Experience Analytics provides one (1) or more of the following outputs:
- Service Experience for a Network Slice: Service Experience for a UE or group of UEs or any UE in a Network Slice;
- Service Experience for an Application: Service Experience for a UE or group of UEs or any UE in an Application or a set of Applications;
- Service Experience for an Edge Application over a UP path: Service experience for a UE or a group UEs or any UE in an Application or a set of Applications over a specific UP path (UPF, DNAI and EC server);
- Service Experience for an Application over a RAT Type or Frequency or both: Service experience for a UE or group of UEs in an Application or a set of Applications over a RAT Type or over a Frequency or both as
defined in Table 6.4.1-1.
- Service Experience for an Application transferring data over a PDU Session: Service experience for a UE or group of UEs or any UE in an Application or a set of Applications transferring data over a PDU Session
with PDU Session parameters i.e. S-NSSAI, DNN, PDU Session Type , SSC mode and optionally an Access Type or with combination of PDU Session parameters such as a list of the tuple (PDU Session Type, SSC
mode) optionally per Access Type.
The 5G CN NWDAF Observed Service Experience can be provided as defined, e.g., individually per UE or Group of UEs, or Globally, Averaged per Application or Averaged across a Set of Applications on a Network
Slice.
The Service "Consumer" can be 5G CN NF (e.g. 5G CN PCF, NSSF, AMF, NEF), AF, or the OAM.
5G CN NWDAF Observed Service Experience related Network Data Analytics
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF Services Data Analytics enablement Exposure for Service Experience Statistics and Predictions
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF Data Analytics on UE Mobility and UE Mobility Predictions
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF Data Analytics on End-to-End (E2E) Data Volume Transfer Time Analytics
5G CN NWDAF collects E2E Data Volume Transfer Time related Input Data from
5GC NFs, OAM and AF. The "Consumer" can either subscribe to Analytics
Notifications (i.e. a Subscribe-Notify Model) or request a Single Notification (i.e.
a Request-Response Model).
The E2E Data Volume Transfer Time refers to a Time Delay for completing the
transmission of a Specific Data Volume from UE to AF, or from AF to UE. The
Data Volume can be the expected or observed Data Volume from UE to AF or
from AF to UE.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5GC to support NWDAF-assisted Policy & QoS control
Use Case #1: NWDAF-assisted QoS enhancement
In 5GS, the QoS Parameters are determined by the 5G CN PCF based on its
knowledge about e.g. AF Requirements, Analytics provided by the NWDAF, etc.
After applying the determined QoS parameters to the Service (as specified in
5GS Architecture 5QI), the PCF may determine whether or not the current QoS
can fully satisfy the Service Requirements based on the Service Experience
Analytics provided by the NWDAF.
If the current 5GS Service QoS cannot satisfy the Service requirements, the 5G
CN PCF updates the QoS Parameters and informs the new parameters to 5G
CN SMF.
Then the 5G CN PCF requires new Service Experience Analytics to check
whether the updated QoS Parameters can satisfy the Service Requirements.
Based on 5GS current QoS Analytics Framework, 5G CN NWDAF requires
several iterations to work out the ideal QoS Parameters.
Using 5G CN NWDAF Analytics knowledge based on Data Collection (DCCF),
5G CN NWDAF assists the 5G CN PCF in determining 5G Service QoS
parameters that can achieve the expected Service Experience Requirements.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5GC to support NWDAF-assisted Policy & QoS control
Use Case #2: Enhancements to QoS Determination with NWDAF Assistance
for the purpose of optimizing the overall Network Performance & Signalling
based on Operator's Policy.
The characteristics of the default QoS is determined by the Subscribed default
Values (for parameters such as 5QI, ARP) which the 5G CN SMF obtains from
the 5G CN UDM. The default QoS rule might be sufficient for basic browsing or
instant messaging over IP, whereas it is not able to satisfy the relatively high
service requirements, i.e. of video streaming applications which require better
QoS treatment.
When the QoS Flows with different requirements from the default flow are
required, Modification to the PDU Session and thereby to establish a new QoS
Flow with the required characteristics is needed. Such modification will result in
significant System-wide Signalling, including NAS Signalling Messages between
the UE and the 5GC, Signalling within 5GC (i.e. signalling between 5G CN SMF,
UPF, PCF Nodes), Signalling between 5GC and RAN, and also the NG-RAN/NR
RRC Messages between the RAN and the UE.
In order to optimize the Network Performance by determining QoS in a more "intelligent" (effective) manner, it would be beneficial for the 5GC to
leverage NWDAF assistance.
When the UE or the Network triggers PDU Session establishment or modification for a new QoS flow with QoS requirements driven by a User or
Service, it would be beneficial if the QoS characteristics are determined by the Network by considering the Predictions and Measurements of some UE
and Network related Information and Service related information (e.g. Service Requirements provided by the AF).
The PDU Session and QoS Flow can be established or modified in a more 'Future-proof' and Multiple-Service-Compatible manner and reduce the
potential modifications of the existing QoS Flow and the corresponding Policy control, e.g. PCC rules.
5GC Functionality enhancements, e.g. of the 5G CN NWDAF, PCF, to enhance the Policy Control and QoS by considering Operator's Policies, will
improve the Network Performance and UE Experience significantly.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

Key Issue #3: NWDAF-assisted Policy Control and QoS enhancement
The NWDAF can gather quite a lot of data from 5GC NFs, AF and OAM and thus may further
assist the PCF in making PCC decisions (which traditionally determine QoS parameters
based on its own data and knowledge as well optional statistics and predictions collected from
the NWDAF).
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G System Application Data Analytics Enablement (ADAE) Functional
Architecture Enhancement to support Application Layer (AL) AI/ML Services
The 5GS Architecture specifies the ADAE Functional Model of with the ML Model
Training and Management Enablement (MTME) Function and the AI/ML
enablement Functional Model Service for supporting Application Layer (AL) AI/ML
Services.
The 5GS ADAE Functional Architecture enhancements with ML Model Training
and Management Enablement Function and AI/ML enablement Service are based
on the 5G Generic Functional Model specified in 5GS Service Enablement
Architecture Layer (SEAL). The 5GS Architecture enhancements are organized
into Functional Entities to describe a Functional Architecture enhancement, which
addresses the support for ML Model Training and Management enablement
Function and AIML enablement aspects for Vertical Applications (VA).
The Figure illustrates the 5GS ADAE On-Network Functional Model with ML
Model Training and Management Enablement (MTME) Function. In the Vertical
Application Layer (VAL), the UE VAL Client communicates with the VAL Server
over VAL-UU Reference Point. VAL-UU supports both Unicast & Multicast delivery
modes. The 5GS ADAE Functional entities with MTME Function on the UE and
the Server are grouped into 5G ADAE Client(s) and 5G ADAE Server(s)
respectively. The 5G ADAE with MTME Function consists of a common Set of
Services (e.g., ML Model provision, (MTME-enhanced) ADAE Client Registration
Management, (MTME-enhanced) ADAE Client member Management, Training
status estimation, ML Model Training execution) and Reference Points. The 5G
ADAE offers its Services to the Vertical Application Layer (VAL)
The 5G ADAE Client(s) communicates with the 5G ADAE Server(s) over the
ADAE-UU Reference Points. The 5G ADAE Client(s) provides the Service Enabler
Layer support Functions to the VAL Client(s) over 5G ADAE-C Reference Points.
The VAL Server(s) communicate with the ADAE Server(s) over the ADAE-S
Reference Points. The ADAE Server(s) communicate with the underlying 3GPP
Network using the respective 3GPP Interfaces specified by the 3GPP Network.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

The 5GS specifies how to maintain a Federation Learning (FL)
Process in FL Execution Phase, including FL Server NWDAF
triggers re-selection, addition, or removal of FL Client NWDAF(s),
discovers new FL Client NWDAF(s) via 5G CN NRF and FL Client
NWDAF(s) joins or leaves Federated Learning process
dynamically.
In the Federated Learning execution phase, FL Server NWDAF
monitors the status changes of FL Client NWDAF(s) and may
reselects FL Client NWDAF(s) based on the updated status,
availability and/or capability.
NOTE1: FL Server NWDAF checks if there is a need to carry
on the FL execution phase and then reselects FL
members for the next iteration if needed.
5GS CN NWDAF Procedures for Maintaining Federated Learning Processes
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5GS AI/ML support for (Vertical) Federated Learning - 1
Federated Learning (FL) is a Machine Learning (ML) Technique that enables multiple FL
Clients to train a Model by exchanging the Parameters instead of exchanging/sharing Local
Data set.
FL Servers perform FL Management Operations by maintaining and updating a Global ML
Model, selecting and managing FL Clients, performing aggregation strategies, scheduling
training in a federated manner, and communicating with FL clients.
FL clients provide various aspects of data for FL operations, such as Data Collection, Data
Preparation, and using Data for training and/or Inferencing while communicating Updates of
ML Models to FL Servers.
There are two (2) types of Federated Learning (FL) defined:
-Horizontal FL and
-Vertical FL Federated Learning.
Horizontal Federated Learning (HFL), or Sample-based FL, is introduced in the scenarios
that Data sets share the same Feature space, but have different Samples.
Vertical Federated Learning (VFL) or Feature-based FL, is applicable to the Cases that two
(2) Data sets share the same Sample space, but differ in Feature space.
In both, Horizontal and Vertical FL Model Parameters from each Local Model Training Function are sent to a Model Aggregator (or FL Collaborator) to
calculate an Aggregate Model. The model aggregator provides updated model parameters to each ML model training entity that each ML model training
entity use to re-train its own model thus allowing every local model training function to have a trained model using data from multiple sources.
An example for vertical FL (VFL) is shown in the Figure hereby:
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF support for VFL -2
In PLMN where multiple 5G CN NWDAFs are deployed, each NWDAF instance may
perform Data collection according to their available Data sources.
Depending on the Analytics ID and the deployment scenario however, the different
NWDAF Instances may share the same Sample space or train on different sample
spaces.
VFL would be beneficial on the former case. Furthermore, in VFL each NWDAF instance
does not collect the same Input Data for the same Analytics ID.
In VFL vendor specific Local Models and Features can be deployed in each participant,
so that it is possible that each participant selects or configures the Local Model to be
used, as such Vendor or Operator specific Local Models and Features, including not
standardized features, are simpler to implement comparing with HFL.
When NWDAF provides observed service experience analytics, as in other analytics that
require input data from the AF, policies in the PLMN and or the AF may prevent raw data
to be exchanged directly between NWDAF and an external AF, as NWDAF is in the
PLMN and the AF is outside the PLMN and the User Data has high Privacy Protection
Needs.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

5G CN NWDAF support for VFL - 3
5G CN NWDAF(s) and AF(s) would collect their Local Training Data, respectively (e.g.
Access Speed, Network Access Delay for NWDAF, Stall Time, Frame Rate for AF).
Two (2) scenarios are available:
Scenario 1: NWDAF initiates VFL Training Process.
Scenario 2: AF initiates VFL training process.
AF and NWDAF can collaborate, e.g. to get DN Performance Analytics.
NOTE 1: This is applicable only if the AF is capable of participating in VFL Procedure as
a Training Entity.
NOTE 2: When an AF initiates the VFL, then only one (1) AF is involved. When the
NWDAF initiates the VFL, then multiple AFs can be involved.
2. 5G System CN NWDAF AI ML use for QoS "Analytics" and Predictions", Vertical Federated Learning (VFL)

Remarks & Questions?

2. 3GPP 5G AIML Enablement (AIMLE) Service Architecture at SYSTEM level

4. 3GPP 5G AIML related activities across all 3GPP WGs - 1

4. 3GPP 5G AIML related activities across all 3GPP WGs - 2 Common AIML Terminology

1. 3GPP all WGs co-ordination for cohesive & integrated Functional Operation for AI ML use in 5G ADVANCED RELEASEs at SYSTEM LEVEL

13
5
3. 3GPP 5G System E2E KPIs - 1

13
6
3. 3GPP 5G System E2E KPIs - 2

13
7
3. 3GPP 5G System E2E KPIs - 3

5G Advanced NG-RAN Direct and Indirect Connection options in 5G Multi-Operator Core Network (5G MOCN)
A 5G Core Network (CN), both, with "Direct" & "Indirect" Connection options between the Shared Access and Core Networks Architecture shall allow Multiple Participating
Operators to share Resources of a Single Shared Network as Radio Access Network (RAN, including Radio Resources. 5G MOCN supports NG-RAN Sharing with or
without Multiple Cell Identity Broadcast as described in 5GS Architecture. 5G MOCN also supports the following sharing Scenarios involving Non-Public Networks (NPNs), i.e.
NG-RAN can be shared by any combination of PLMNs, PNI-NPNs (with CAG), & SNPNs (each identified by PLMN ID & NID). In all NPN Sharing Scenarios, each Cell Identity
as specified in 5G NG-RAN, is associated with one of the following configuration options: A) One (1) or Multiple SNPNs; B) One (1) or Multiple PNI-NPNs (with CAG); or C) One
(1) or Multiple PLMNs only. This allows the assignment of Multiple Cell Identities (Cell-IDs) to a Cell & also allows the Cell-IDs to be independently assigned, i.e. without need for
coordination, by the Network sharing Partners, between PLMNs &/or NPNs. Different PLMN IDs (or combinations of PLMN ID & NID) can also point to the same 5GC. When same
5GC supports multiple SNPNs (identified by PLMN ID & NID), it is up to the Operator's Policy whether they are used as "equivalent" SNPNs for a UE. A 5G Network may
utilise the PLMN-assigned UE Radio Capability ID, without involving the UE, e.g. for use with legacy UEs. RACS (UE RAdio Capability Signalling optimization) works
by assigning an Identifier to represent a Set of UE Radio Capabilities. A UE Radio Capability ID can be either A) UE Manufacturer-assigned or B) PLMN-assigned, as specified
in 5GS Architecture. The UE Radio Capability ID is an alternative to the Signalling of the UE Radio Capability information over the Radio Interface, within NG-RAN, from NG-RAN
to E-UTRAN, from AMF to NG-RAN & between CN Nodes supporting RACS. A NG-RAN which supports RACS (UE RAdio Capability Signalling optimization) can be configured to
operate with one (1) of two (2) Modes of Operation, when providing the UE Radio Capabilities to the 5G CN AMF when the NG-RAN executes a UE Radio Capability Enquiry
Procedure to retrieve UE Radio Capabilities from the UE. Mode of Operation A): The NG-RAN provides to the 5G CN AMF both RAN formats (NR and LTE E-UTRAN
Radio Capabilities formats). The NG-RAN derives one (1) of the UE Radio Capability Formats using local transcoding of the other format it receives from the UE &
extracts the E-UTRAN UE Radio Capability for Paging & NR UE Radio Capability for Paging from the UE Radio Capabilities. Mode of Operation B): The NG-RAN provides to
the 5G CN AMF the NR Radio Capabilities format only. In a PLMN supporting RACS only in 5GS, Mode of Operation B shall be configured. When Users access through a Shared
Network, it is still expected the Hosted Services to be available for Subscribers visiting Hosted Services via Shared Networks. As shown in the Figure below, OP1 is a Hosting NG-
RAN Operator. The CN of OP2, as the Participating Operator, does not have Direct Connection with OP1’s Shared NG-RAN. There is Connection between the OP1’s CN & OP2’s
CN. OP2 provides Hosted Services for Subscribers (HSSs) in the Local Area (LA) 1 & 2. OP2 wishes to provide HSSs accessing through OP1’s & OP2’s Network in the LA 1.

13
9

Remarks & Questions?

Ike A Common MDA for 5GS Telecom and O-RAN Alliance SMO Rev PA06 Feb 2025.pdf

About This Presentation

Slide Content

Tags

Categories

Download

Quick Actions

Statistics

Related Slideshows

Ike A Common MDA for 5GS Telecom and O-RAN Alliance SMO Rev PA06 Feb 2025.pdf

About This Presentation

Slide Content

Slide 1

Slide 2

Slide 3

Slide 4

Slide 5

Slide 6

Slide 7

Slide 8

Slide 9

Slide 10

Slide 11

Slide 12

Slide 13

Slide 14

Slide 15

Slide 16

Slide 17

Slide 18

Slide 19

Slide 20

Slide 21

Slide 22

Slide 23

Slide 24

Slide 25

Slide 26

Slide 27

Slide 28

Slide 29

Slide 31

Slide 32

Slide 33

Slide 34

Slide 35

Slide 36

Slide 37

Slide 38

Slide 39

Slide 40

Slide 41

Slide 42

Slide 43

Slide 44

Slide 45

Slide 46

Slide 50

Slide 51

Slide 52

Slide 53

Slide 54

Slide 55

Slide 56

Slide 57

Slide 58

Slide 59

Slide 60

Slide 61

Slide 62

Slide 63

Slide 64

Slide 65

Slide 66

Slide 67

Slide 68

Slide 69

Slide 70

Slide 71

Slide 72

Slide 73

Slide 74

Slide 75

Slide 76

Slide 77

Slide 78

Slide 79

Slide 80

Slide 81