hash stringlengths 32 32 | doc_id stringlengths 5 12 | section stringlengths 5 1.47k | content stringlengths 0 6.67M |
|---|---|---|---|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4 ML model deployment
| Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4.1 ML model transfer delivery to UE
| Editor’s note: The content in this clause remains for further discussion and depends on further correspondence and confirmation from the relevant RAN WGs.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4.1.1 Description
| RAN1 and RAN2 started in Rel-18 a joint study on AI/ML for NR air interface (see FS_NR_AIML_air [5]) whose outcomes were reported in TR 38.843 [3], approved in RAN#102 (Dec. 2023). One of the features in scope of this study was ML model transfer/delivery to UE. The clause 7.2.1.4 of TR 38.843 [3] documented several potential solutions for this feature. One of these solutions is Solution 4b, which involves OAM. For Solution 4b, TR 38.843 [3] quotes the following:
“Note: For Solution 4b, RAN2 discussed the following two solutions but did not study or analyse their feasibility:
- OAM can transfer/deliver AI/ML models to UE via OAM→RAN→UE, where Control Plane (CP) signalling is used for RAN→UE.
- OAM can transfer/deliver AI/ML models to UE via OAM→UE, e.g., via IP tunnel.”
Based on this quote, SA5 decided to include this feature in the scope of FS_AIML_MGT_Ph2 [6], a Rel-19 study whose outcomes were reported in TR 28.858 [7]. However, after investigating the two use cases defined for solution 4b, SA5 did not draw any definite conclusions. Clause 5.3.2.5 of TR 28.858 [7] quoted the following:
“[…] further enhancements to ML model loading capabilities defined by SA5 may still be needed in the option of OAM transferring/delivering AI/ML models to UE via OAM→RAN→UE, where Control Plane (CP) signalling is used for RAN→UE and should be investigated in the normative work. Furthermore, the option of OAM transferring/delivering AI/ML models to UE via OAM →UE, e.g. via IP tunnel is FFS and needs to be investigated in SA5 once the requirements are clear from RAN working groups”.
Despite clause 6 of TR 28.858 [7] recommended SA5 to specify in Rel-19 normative phase the AI/ML management capabilities for the ML model delivery/transfer feature as defined by RAN1/2, this feature was never revisited. The reason was the lack of progress in RAN1/2 for this feature during Rel-19. On one side, RAN2 concluded their work in Rel-18 with the documentation of potential solutions in clause 7.2.1.4 of TR 38.843 [3], and did not follow up the work in Rel-19. On the other side, RAN1 was tasked to analyse on the need and feasibility of these potential solutions. After concluding this analysis, RAN1 could not reach a consensus in Rel-19 on the need of a standardized solution for the ML model delivery/transfer feature. This is captured in the RAN1#120 meeting Chair notes [8], section 9.1.4.2.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4.1.2 Potential requirements
| None.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4.1.3 Possible solutions
| SA5 should look at conclusion documented in RAN1#120 meeting Chair notes [8] section 9.1.4.2, and should not discuss solution 4b for now, given that RAN has not provided any guidance to SA5 on the need to standardize a solution for this feature.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.4.1.4 Possible solutions evaluation
| With no further action required from RAN, the analysis of the two options defined for Solution 4b would remain as follows:
The option “OAM can transfer/deliver AI/ML models to UE via OAM→UE, e.g., via IP tunnel” in Solution 4b is not feasible, given that the 3GPP management system does not have a direct interface with the UE.
The option “OAM can transfer/deliver AI/ML models to UE via OAM→RAN→UE, where Control Plane (CP) signalling is used for RAN→UE” in Solution 4b is not feasible, given that RAN2 has not introduced any CP signalling to support this in Rel-19, and that is not in scope of Rel-20 in RAN2.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 5.5 AI/ML inference
| Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6 AI/ML sustainability
| |
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0 Sustainable aspects of ML model training and inference
| |
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.1 Description
| ML model training and AI/ML inference can require substantial computing power and therefore significant energy.
For sustainability goals in AI/ML management, it is valuable for the 3GPP management system to reflect the type of energy supplied to the managed entities that are used for ML model training and AI/ML inference (for example, renewable and/or carbon related information). The operator handles this energy related information (e.g., see EnergySupplyInfo in TS 28.310 [9]) in the 3GPP management system, based on knowledge of the infrastructure it manages. In particular:
- the operator obtains energy related information typically based on data from external energy suppliers; the operator sets, maintains and updates the association of energy related information to managed entities that support ML model training and inference, by configuration of the Energy Information defined in TS 28.310 [9].
- the operator defines access control rules/policies to determine what managed entities (e.g. NFs, AI/ML functions) can gain access to this information.
In the context of ML model training and AI/ML inference, the MnS producer can read the operator-configured energy related information, subject to operator-defined access control. By having access to this information, the MnS producer can support receiving training/inference requests containing energy related information. These requests are triggered by the operator, who performs the role of MnS consumer.
In the context of ML model training, the MnS producer can also expose the operator-configured energy related information to other training functions, for the case these training functions need to know this information for selection criteria. This happens for example in Federated Learning (where FL client, acting as MnS producer, exposes energy related information to other training functions acting as FL servers, for them to manage FL lifecycle, e.g. choose FL clients based on energy criteria) and Distributed Training.
The MnS producer reads operator-configured information and applies it when selecting functions for model training or inference. The accuracy of this information depend on operator configuration and external agreements, and are outside the scope of MnS producer verification.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.2 Potential requirements
| REQ-ML_SUST-01. The 3GPP management system should enable the operator to configure and update the energy related information for the managed entities representing an ML model training function.
REQ-ML_SUST-02. The 3GPP management system should expose the energy information of the managed entities representing an ML model training function to authorised MnS consumers.
REQ-ML_SUST-03. The 3GPP management system should enable the operator to configure and update the energy information for the managed entities representing AI/ML inference.
REQ-ML_SUST-04. ML Training Function should enable an MnS consumer to request for ML model training with consideration to energy related information.
REQ-ML_SUST-05. ML Training Function should be able to adapt its training to the energy related information updates, according to the conditions provided by MnS consumer which requests the ML model training.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.3 Possible solutions
| It is proposed one solution consisting of two parts: one impacting TS 28.310 [9] (see clause 6.0.3.1), and other impacting TS 28.105 [4] (see clause 6.0.3.2).
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.3.1 Update Energy Information NRM fragment
| It is proposed to update the Energy Information NRM fragment represented in 8.2.1 of TS 28.310 [9], by adding in ManagedEntity the following:
- The MLTrainingFunction IOC (see clause 7.3a.1.2.1 in TS 28.105 [4])
- The list of IOCs represented by AIMLSupportedFunction (see clause 7.3a.4.1.1 in TS 28.105 [4])
These additions require importing IOCs from 3GPP TS 28.105 [4].
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.3.2 Update MLTrainingFunction
| It is proposed to add a new attribute to MLTrainingFunction IOC: “energyInfoGroupRef”. This attribute allows specifying the DN of the energyInfoGroup MOI representing the operator-configured energy information.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.0.4 Possible solutions evaluation
| The solution described in clause 6.0.3 is feasible.
The part of the solution described in clause 6.0.3.1 allows satisfying REQ-ML_SUST-01 and REQ-ML_SUST-03.
The part of the solution described in clause 6.0.3.2 allows satisfying REQ-ML-SUST-02.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.1 Sustainability for ML training
| Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 6.2 Sustainability for AI/ML inference
| Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 7 Registration and discovery for AI/ML
| Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 8 Relation with other management capabilities
|
Editor’s note: Similar clause structure as in subclause 5.1 will be adopted.
|
b4e3b34fa206f18d0c88d81247788555 | 28.882 | 9 Conclusion and recommendations
|
Annex A (informative):
Change history
Change history
Date
Meeting
TDoc
CR
Rev
Cat
Subject/Comment
New version
2025-08
SA5#162
Initial skeleton
0.0.0
2025-10
SA5#163
S5-254664
S5-254666
S5-254667
S5-254744
pCR
pCR
pCR
pCR
pCR TR 28.882 initial ToC
Pseudo-CR on TR 28.882 add Management support to UE-Side model training use case
Pseudo-CR on TR 28.882 add Management support to NW-Side model training use case
Pseudo-CR on TR 28.882 Add New Use Case on Management of Vertical Federated Learning
0.1.0
2025-11
SA5#164
S5-255505
S5-255506
S5-255515
S5-255518
S5-255519
S5-255687
pCR
pCR
pCR
pCR
pCR
pCR
Pseudo-CR TR 28.882 Use case on ML model transfer delivery to UE
Pseudo-CR on TR 28.882 add Management support use case and requirement for two-sided model training
Pseudo-CR TR 28.882 Add Solution for Management of Vertical Federated Learning
pCR on Rel-20 TR 28.882 Adding Enhancements on LCM of Federated Learning
Pseudo-CR TR 28.882 Sustainability aspects of ML model training and inference
Rel-20 Pseudo-CR TR 28.882 Use case on enhancing Reinforcement Learning with performance targets
0.2.0
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 1 Scope
| The present document describes use cases, potential requirements, and potential solutions aimed at enhancing the Network Digital Twin (NDT) defined in TS 28.561, with aspects on interaction and collaboration between NDT and network functions/automation functions, multiple NDT collaborations, data collection requirements to support NDTs and any other new use cases. It also presents conclusions and recommendations regarding the next steps in the 3GPP standardization process.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 2 References
| The following documents contain provisions which, through reference in this text, constitute provisions of the present document.
- References are either specific (identified by date of publication, edition number, version number, etc.) or non‑specific.
- For a specific reference, subsequent revisions do not apply.
- For a non-specific reference, the latest version applies. In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.
[1] 3GPP TR 21.905: "Vocabulary for 3GPP Specifications".
[2] 3GPP TS 28.312: " Management and orchestration; Intent driven management services for mobile networks".
[3] 3GPP TS 28.561: " Management and orchestration; Management aspects of Network Digital Twins".
[4] 3GPP TS 28.532: "Management and orchestration; Generic management services".
[5] 3GPP TS 38.300: "NR; NR and NG-RAN Overall description; Stage-2"
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 3 Definitions of terms, symbols and abbreviations
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 3.1 Terms
| For the purposes of the present document, the terms given in TR 21.905 [1] and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in TR 21.905 [1].
example: text used to clarify abstract rules by applying them literally.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 3.2 Symbols
| For the purposes of the present document, the following symbols apply:
<symbol> <Explanation>
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 3.3 Abbreviations
| For the purposes of the present document, the abbreviations given in TR 21.905 [1] and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in TR 21.905 [1].
<ABBREVIATION> <Expansion>
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 4 Concepts and background
| Editor’s note: This clause provides a description of concepts and background.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5 Use cases
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.1 Use Case #1: NDT support intent pre-evaluation
| 5.1.1 Description
TS 28.312 [1] introduces the intent pre-evaluation which includes Intent Feasibility check and Intent Exploration. Intent feasibility check enables the MnS consumer to check if the proposed intent can be supported by the MnS producer. Intent exploration enables the MnS consumer and the MnS producer to find the intent for fulfilment that is best aligned with MnS producer's capabilities. NDTF can be used to support for intent pre-evaluation.
The relation between NDTF and Intent Handling Function follows the relation between NDTFs and network automation functions defined in TS 28.561 clause 4.3. An adaption figure based on Figure 4.3-1 in TS 28.561 is shown below.
Figure 5.X.1-1: Relation between NDTF and Intent Handling Function
In the scenario of radio network/service intent pre-evaluation, the MnS consumer may request to obtain the best values for a given target or context, e.g., the number of terminal devices. NDTF can support intent handling function to evaluate the number of terminal devices (e.g., UE) given certain simulation conditions, such as intent objects scope (e.g., area scope, cell lists, civic address, etc), intent object type (e.g., radio service, radio network, etc). IHF requests NDTF to simulate the network performance under various conditions, NDTF simulates the network performance and reports to IHF with simulation results.
This use case proposes to enhance the description related to network automation in TS 28.561 to capture the relationship between NDT and IHF and specify that NDT can support network automation capability including intent pre-evaluation.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.1.2 Potential requirements
| No new requirements
5.1.3 Potential solutions
This solution proposes following updates in TS 28.561[3]:
1. In clause 4.3, add Intent Handling function as one additional example of network automation functions.
2. In clause 5.2.2.1, add the text below:
“NDT can support for Intent Handling Function enabling intent handling capability, e.g., feasibility check and exploration capability. For example, in the scenario of radio network/service intent feasibility check and exploration, the MnS consumer may request to obtain the best values for a given target or context, e.g., the number of terminal devices. NDTF can support intent handling function to evaluate the number of terminal devices (e.g., UE) given certain simulation conditions, such as intent objects scope (e.g., area scope, cell lists, civic address, etc), intent object type (e.g., radio service, radio network, etc). IHF requests NDTF to simulate the network performance under various conditions, NDTF simulates the network performance and reports to IHF with simulation results.”
5.1.4 Evaluation of potential solutions
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.2 Use Case #2: Improvement of data generation
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.2.1 Description
| TS 28.561 clause 5.4.2.2 [3] introduces the data generation capability provided by NDT to enable ML training. For different ML models, the preferences of its training data are different, which can vary in the dimension of data source object, data type, data quantity. The data source object is used to specify the simulated object from which the synthetic ML training data is collected. It is the subset of the NDT synchronization objects. The data quantity is used to specify how many data needs to be reported by NDT for ML training.
The existing solution of NDT NRMs can support the data generation by specifying data source objects and data type to be collected. The data source objects can be specified by the managedEntitiesScope defined in ScopeDefinition <<choice>> which represents the synchronization scope. The data type can be specified by the simulationData defined in SimulationDataDescriptor <<dataType>>. However, the existing solution cannot satisfy the data generation scenario where the requested data source objects are the subset of the synchronized network objects with specific data quantity.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.2.2 Potential requirements
| REQ-NDTDG-01: The 3GPP management system should support a capability to allow an authorized MnS consumer to express data generation preferences on data source object, data type, and data quantity.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.2.3 Potential solutions
| This solution aims to enhance NDT NRMs as shown below:
1. Enhancement for NDTJob IOC
- Add new attribute “nDTDataGenObject” which represents the data source object from which the synthetic data is collected. This attribute allows the MnS consumer to express data generation preferences on data source objects.
- Add new attribute “dataQuantity” which represents the requested synthetic quantity of data used for ML training. This attribute allows the MnS consumer to express data generation preferences on data quantity. This attribute can either be an integer which represent the collection times for the synthetic data, or be a combination of time window and frequency which implicitly represents the data quantity.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.2.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.3 Use case #3: Collaborate with ML training Producer to generate data
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.3.1 Description
| In 3GPP TS 28.561 [3], the existing use case and requirements for using NDT to generate ML training data is described in clause 5.4.2.2. However, using NDT alone to generate data may be insufficient to support the following scenarios:
- Due to the coverage limitations of the physical network simulated by NDT, some extreme scenario data (such as sudden traffic peaks or unforeseen equipment failures) may not be generated.
- NDT-based data generation depends on complex procedures such as simulating network topology and device interactions, making it time-consuming when generating large volumes of data.
Therefore, for scenarios requiring extreme data generation or large-scale data generation, it is considered to introduce AI-based data generation models, generated by the ML training Producer, into NDT to enable rapid, batch, and comprehensive data generation.
As shown in the Figure 5.3.1-1, the MnS Consumer can request the MnS producer to create an NDT instance for generating data with an indication of simulation object, data type, and data requirements, etc. Data requirements may specify large-scale data, extreme data requirements. The MnS producer creates an NDT instance based on the request and sends a response to the MnS consumer. The MnS producer can act as an ML training consumer to send a request to the ML training producer for generating a data generation model. Subsequently, the MnS producer executes simulation based on the NDT instance to obtain simulation data (e.g.,the generated UE throughput data), which is then sent to the ML training producer. This simulation data is used as training data to update and train the data generation model. The MnS producer can act as an ML inference function to receive the updated model from the ML training producer, execute it to obtain the final generated data, and send this data to the MnS consumer.
Figure 5.3.1-1 Collaborate with ML training Producer to generate data
Through this method, after the ML training producer completes ML model training and updates based on the initial NDT simulation data, it only needs to perform AI inference for subsequent data generation, which can reduce certain resource consumption.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.3.2 Potential requirements
| REQ-NDTDG-AI-1: The 3GPP management system should support a capability that enables an authorized MnS consumer to obtain the enabler information of the NDT data generation.
5.3.3 Potential solutions
5.3.4 Evaluation of potential solutions
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.4 Use Case #4: Enhancement for multiple NDT collaborations
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.4.1 Description
| In 3GPP TS 28.561 [3], the existing use case and requirements for collaboration between NDTs are described in clause 5.5.2.1 and the requirement in clause 5.5.3. The 3GPP management system should support a capability enabling an authorized MnS consumer to configure the relationship between NDTs during simulation/emulation. However, there are no specific solutions to support this scenario, and there are no further details regarding the collaboration of multiple NDTs.
Therefore, the scenario needs the further investigate, a single NDT Function might not be able to fulfil a task by itself and may depend on or need to use the service or outputs of another NDT Function during the simulation/emulation activity. This require the 3GPP management to support the capabilities and report the relationships between NDTs regarding the collaboration of multiple NDTs.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.4.2 Potential requirements
| REQ-NDT-Colla-1: The 3GPP management system should support a capability that enables an authorized MnS consumer to request a report on the relationships between NDTs regarding the collaboration of multiple NDTs.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.4.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.4.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.5 Use Case #5: Enhancement on NDT reporting method
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.5.1 Description
| In TS 28.561 [3], the reporting method of NDT output is notification based reporting. The NDT MnS consumer receives an NDT report from NDT producer by invoking generic provisioning management service operations and notifications in TS 28.532 [4]. However, different NDT job may have different requirement on reporting method. For example, in the scenario of signalling storm, it is more appropriate to use streaming data reporting service to timely report the simulation/emulation information on potential network impact.
It is proposed to enable the NDT MnS consumer to select the reporting method of simulation/emulation output based on different requirement. In addition to notification based reporting, streaming data reporting service also need to be supported for NDT output reporting.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.5.2 Potential requirements
| REQ-NDTRM-01: The 3GPP management system should support a capability to allow an authorized MnS consumer to request NDT report to be provided by streaming data reporting service.
REQ-NDTRM-02: The 3GPP management system should support a capability for NDT MnS producer to support delivery of NDT report by streaming data reporting service.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.5.3 Potential solutions
| It is proposed to make the following changes to TS 28.561 [3] clause 6.2.1.3.2:
- update the NDTJob IOC attribute to add a new optional attribute reportingMethod, which can be used to indicate the required reporting method of NDT report, the allowed values of reportingMethod can include streaming and notification. The new attribute can only be used when the requested NDTFunction supports the capability that allow a MnS consumer to indicate the reporting method.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.5.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.6 Use Case #6: Capability Discovery of NDT in NDT Collaboration
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.6.1 Description
| NDT collaboration among different NDTFunctions located in different domain is critical to enable end to end use cases, In Rel-19 the supported domain of NDT is represented by NDTFunctionScope attribute. Then, the NDTFunction that triggers the collaboration can involve NDT(s) that supports modelling of different domain.
When simulating end-to-end user experience data generation, the consumer may want to specify diverse traffic model and UE distribution for different scenarios. For example, major event such as sporting events, the UE distribution and user’s communication behaviour is quite different, in which the uplink traffic will increase significantly. Therefore, if consumers want to trigger NDT collaboration for E2E user experience data generation, they should know which candidate NDT supports the simulation of different traffic. Therefore, NDTFunction needs further enhancement to expose more detailed capability information (e.g., capability of supporting UE distribution model and traffic model). Then, the consumers can accurately identify and select appropriate collaboration participants as NDT component, ensure cross-domain traffic consistency, and effectively validate end-to-end scenario continuity.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.6.2 Potential requirements
| REQ- NDTAUT-01
The 3GPP management system should support a capability enabling authorized MnS Consumer to discover detailed capability of candidate NDT components for NDT collaboration.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.6.3 Potential solutions
| Add a new the GenericModellingCapabilites <<dataType>> with the following new attributes to support modelling capability discovery of NDT in NDT collaboration:
• EnviromentModelling, it represents the NDTFunction supports to model the network topology and network infrastructure.
• UETrafficModelling, an ENUM, which indicates the capability of traffic modelling supported by the NDTFunction and may include the following values:
1) FixedUEModel, it represents that the NDTFunction supports to model the location of the UE. It means that the NDT can support to model the traffic of a single UE. The information of specific UE is unknown for the NDFFunction.
2) MovingUEModel, it represents that the NDTFunction supports to model the trajectory of the moving UE. It only means that the NDT can support to model the traffic of a moving UE. The information of specific UE is unknown for the NDTFunction.
3) DistributedUEsModel, it represents that the NDTFunction supports to model the distribution of UEs.
• ServiceModelling, an ENUM, which represents the service that the NDTFunction supports, e.g., URLLC,
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.6.4 Evaluation of potential solutions
| TBD
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.7 Use Case #7 Defining the Lifecycle and Runtime Behaviour of NDT Jobs
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.7.1 Description
| The NDTJob lifecycle is not clear, there are some items open to interpretation and some items missing which are described below.
Issue#1: The MnS Consumer should have a method to know which configuration of NDTJob has produced a given report. In the current specification NDTReport refer to the NDT job (i.e NDTJobRef). In the existing specification, it is possible to modify the NDTJob which may result in the different NDTReport. This will result in loosing the link between produced NDT Report and the original NDT Job. The MnS Consumer would benefit from understanding the implications of reconfiguring the NDTJob. At the moment, the NDTReports refer to the DN of the NDTJob – if a NDTJob is reconfigured, this means the same DN is applied to each report, even if the simulation has changed. Likewise, there is no clear procedure described in Clause 6.4, despite the “Modify” use-case being possible for the NDTJob Instances.
Excerpt: Table 6.2.1.3.8.2-1 describes the attributes associated with the NDTReport <<IOC>>
Excerpt: From 6.3 Attribute definitions, the explanation of the ndtJobRef of the NDTReport <<IOC>> refers to
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.7.2 Potential requirements
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.7.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.7.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.8 Use Case #8: Using external data for NDT modelling
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.8.1 Description
| As defined in TS 28.561[3], Network Digital Twin (NDT) is used as a replica of a mobile network, in order to learn how an actual mobile network would behave in certain scenarios, without causing any changes to the actual mobile network. To provide meaningful results, an NDT needs to model the behaviour of the mobile network, so that the result of the operations on the virtual replica is a good approximation to the result of similar operations on the actual network. The accuracy of the approximation relies on how much the NDT modelling mimics the live network, which is further impacted by the data collected for NDT modelling.
Especially in the case of RAN domain NDT, the RAN communication performance may be additionally impacted by the environment, such as environmental buildings, weather, etc. NDT can make use of such external data for NDT modelling to provide a more accurate approximation of the physical network. The existing solution of NDT NRMs can specify the NDT job synchronization scope which is either the DN(s) of managed entities or the geographical area information. How the external data can by specified by NDT MnS consumer is to be addressed.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.8.2 Potential requirements
| REQ-NDTDG-01: The 3GPP management system should support a capability to allow an authorized MnS consumer to specify external data information used for NDT modelling.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.8.3 Potential solutions
| TS 28.622 defines the ExternalDataType IOC which specifies a type of external management data and the associated meta data. This solution aims to enhance NDTJob IOC and NDTFunction IOC by reusing ExternalDataType IOC to carry the NDT MnS consumer’s external data preference for an NDT job and show the supported external data of NDT Function used for NDT modelling. Details are as shown below:
1. Enhancement for NDTJob IOC
- Add new attribute “externalDataTypeRefList” representing the NDT Job is associated with one or more other ExternalDataType MOIs that are used for NDT modelling.
- Add new attribute “supportedExternalDataList” for NDTFunction IOC which represents the external data supported by the NDTFunction for NDT modelling.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.8.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.9 Use Case #9: NDT for Non-Terrestrial Network (NTN) Performance and Optimization Evaluation
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.9.1 Description
| According to 3GPP 38.300 [5] a Non-Terrestrial Network (NTN) is defined as an NG-RAN consisting of gNBs, which provide non-terrestrial NR access to UEs by means of an NTN payload embarked on an airborne or space-borne NTN vehicle and an NTN Gateway. NTN Gateway is an earth station located at the surface of the earth, providing connectivity to the NTN payload using the feeder link which is a wireless transport link between the NTN Gateway and the NTN payload. NTN payload is a network node, embarked on board a satellite or high-altitude platform station, providing connectivity functions, between the service link and the feeder link.
Compared with terrestrial networks, NTNs exhibit unique characteristics such as high mobility of nodes, dynamic topologies, long propagation delays, and variable link conditions. These aspects make planning, optimization, and management challenging for mobile network operators.
This use case proposes using Network Digital Twin (NDT) technology to simulate, analyze, and optimize the behavior of NTNs. For example, the NDT can create a virtual representation of the satellite constellation, NTN Gateways, and user terminals to predict network conditions and support decision-making for service continuity and QoS/QoE management.
The NDT framework for NTN can be used for example to:
• Predict and mitigate handover failures during satellite transitions for NTN entities' constellations.
• Optimize beam management and power allocation based on user distribution and traffic demand.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.9.2 Potential requirements
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.9.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.9.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.10 Use Case #10: Clarification of NDTJob Modification Behaviour
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.10.1 Description
| The MnS Consumer requires clarity on the conditions and procedures for modifying an existing NDTJob instance. At the moment, the specification leaves this open to interpretation, even though there are clear limitations associated with different implementation techniques for NDTJobs. For instance, if an NDTJob is implemented as a script that performs calculations based on input from the MnS Consumer and the replicated network, modifying such a script during execution would effectively mean rerunning the script with new parameters as resumption of certain scripts would be impossible. On the other hand, if the implementation is utilising a technology which can enable modification during runtime, then modification during the initial stages—or at specific points within the runtime—may still be possible.
Currently, the specification does not allow for clear modelling of the different potential stages executed during an NDTJob. As a result, modification of an NDTJob will be inconsistent across implementations, making MnS Consumer integration bespoke to specific NDTs.
The current specification does not define a procedure or consequences of the modification of an NDTJob. TS 28.561 [3] Clause 4.4 (“NDT life-cycle management”) states that “the NDT job instance can be configured by the MnS Consumer at any time”, yet no guidance is provided on permissible modification timing, allowable attribute changes, or the expected system behaviour when an NDTJob is already under execution.
This ambiguity raises questions such as:
• When, during the NDTJob lifecycle, modification requests are acceptable.
• Whether modification during execution affects job consistency or results.
• How the MnS Producer should report modification status or failures.
A clearer model describing NDTJob state transitions and management conditions would help ensure consistent interpretation across implementations. Such a model could make it explicit when an NDTJob is in a state that permits modification and when it is not.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.10.2 Potential requirements
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.10.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.10.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.11 Use Case #11: Create and Execute NDT Job
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.11.1 Description
| The MnS Consumer requires clarity on whether the creation and execution of an NDTJob are distinct phases or a single combined procedure.
By having creation of an NDTJob separate to the execution of the NDTJob, it means certain MnS Consumers can define the NDTJob, and other MnS Consumers could execute the NDT jobs. This is beneficial to the MnS Consumers as users with different roles can create the NDTJob such as subject matter experts, then a different set of users can execute the simulation, which creates a clear separation of responsibilities.
In the current specification, the relationship between these phases is inconsistent across clauses:
• Clause 4.4 (NDT Life-cycle Management) describes “NDT job instantiation” as including both creation and execution within a single step, suggesting automatic execution upon creation.
• Clause 6.4 (Procedures for NDT operations) depicts separate interactions between the MnS Consumer and MnS Producer for creation and execution, implying that execution follows a distinct request.
• Clause 7.1 (RESTful HTTP-based solution set) defines only the createMOI operation for NDTJob creation, with no corresponding operation for initiating execution.
This discrepancy makes it unclear how an MnS Consumer can prepare an NDTJob configuration and decide when to initiate its execution.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.11.2 Potential requirements
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.11.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.11.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.12 Use Case #12: Clarification of Suspension and Resumption Capabilities for NDTJobs
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.12.1 Description
| The MnS Consumer requires clarity on whether an NDTJob can be suspended and subsequently resumed, when such capabilities are supported by the NDT MnS Producer.
The MnS Consumer should have the capability to suspend NDTJobs such as an optimization activity in order to free up resources to prioritize different NDTJobs such as those surrounding faults, otherwise the NDT MnS Producer may not have the available capacity to execute multiple NDTJobs in parallel.
The current specification references the suspension and resumption of NDTJobs in multiple clauses but provides no corresponding procedures, state descriptions, or information model attributes. This creates ambiguity regarding:
• Whether suspension and resumption are supported capabilities or illustrative examples;
• How the MnS Consumer is expected to request suspension or resumption;
• What the expected lifecycle state transitions are following suspension or resumption;
• How results or reports are handled when an NDTJob is suspended.
The capability to suspend and resume NDTJobs should remain optional, dependent on MnS Producer implementation. However, explicit clarification in the specification would help ensure consistent behaviour and interface expectations across implementations.
|
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.12.2 Potential requirements
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.12.3 Potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 5.12.4 Evaluation of potential solutions
| |
282fa1049c2866220729ce3d5d76aa25 | 28.883 | 6 Conclusions and Recommendations
| 6.X Use case #<X>: <use case title>
Editor's note: This clause provides conclusions and recommendations for the corresponding use case.
Annex <X> (informative):
Change history
Change history
Date
Meeting
TDoc
CR
Rev
Cat
Subject/Comment
New version
2025-08
SA5#162
Initial skeleton
V0.0.0
2025-10
SA5#163
1. S5-254290
2. S5-254670
3. S5-254842
4. S5-254672
5. S5-254843
6. S5-254396
7. S5-254674
8. S5-254733
1. Rel-20 pCR TR 28.883 Add structure proposal
2. Rel-20 pCR TR 28.883 Use case on NDT support intent pre-evaluation
3. Rel-20 pCR TR 28.883 Improvement of data generation
4. pCR TR 28.883 Add a use case of NDT data generation
5. pCR TR 28.883 Add use case and requirements on enhancement for multiple NDT collaborations
6. Rel-20 pCR TR 28.883 Enhancement on NDT reporting method
7. Pseudo-CR on TR 28.883 Add New Use Case on Capability Discovery of NDT in NDT Collaboration
8. Rel-20 pCR TR 28.883 Defining the Lifecycle and Runtime Behaviour of NDT Jobs
V0.1.0
2025-11
SA5#164
1. S5-255521
2. S5-255522
3. S5-255523
4. S5-255524
5. S5-255525
6. S5-255526
7. S5-255527
8. S5-255528
9. S5-255529
10. S5-255531
11. S5-255532
1. Rel-20 pCR TR 28.883 Add introduction and scope
2. Rel-20 pCR TR 28.883 Solution for NDT supporting intent feasibility check and exploration
3. Pseudo-CR TR 28.883 Add Solution for Capability Discovery of NDT in NDT Collaboration
4. Rel-20 pCR TR 28.883 Solution for improvement of data generation
5. pCR TR 28.883 Add requirements for NDT data generation
6. Rel-20 pCR TR 28.883 Using external data for NDT modelling
7. Pseudo-CR on TR28.883 Use case about NDT for NTN
8. pCR on TR 28.883 Add solution of NDT reporting method
9. Rel-20 pCR TR 28.883 Clarification of NDTJob Modification Behaviour
10. Rel-20 pCR TR 28.883 Create and Execute NDT Job
11. Rel-20 pCR TR 28.883 Clarification of Suspension and Resumption Capabilities for NDTJobs
V0.2.0
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 1 Scope
| The present document describes use cases, potential requirements, and potential solutions aimed at enhancing the Service Based Management Architecture (SBMA) in the context of 5G Advanced (5GA). It also presents conclusions and recommendations regarding the next steps in the standardization process.
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 2 References
| The following documents contain provisions which, through reference in this text, constitute provisions of the present document.
- References are either specific (identified by date of publication, edition number, version number, etc.) or non‑specific.
- For a specific reference, subsequent revisions do not apply.
- For a non-specific reference, the latest version applies. In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.
[1] 3GPP TR 21.905: "Vocabulary for 3GPP Specifications".
[2] 3GPP TS 28.533: "Management and orchestration; Architecture framework".
[3] 3GPP TS 28.532: " Management and orchestration; Generic management services".
[4] 3GPP TS 28.537: " Management and orchestration; Management capabilities".
[5] 3GPP TS 28.552: " Management and orchestration; 5G performance measurements".
[6] 3GPP TS 28.554: " Management and orchestration; 5G end to end Key Performance Indicators (KPIs)".
[7] 3GPP TS 32.423: " Telecommunication management; Subscriber and equipment trace: Trace data definition and management".
[8] https://datatracker.ietf.org/doc/html/rfc6455
[9] https://websocket.org/guides/websocket-protocol/
[10] 3GPP TS 28.111: "Management and orchestration; Fault management (FM)".
[11] 3GPP TS 32.531: "Telecommunication management; Software management (SwM); Concepts and Integration Reference Point (IRP) Requirements".
[12] 3GPP TS 32.532: "Telecommunication management; Software management (SwM); Integration Reference Point (IRP); Information Service (IS)".
[13] 3GPP TS 32.533: "Telecommunication management; Software management (SwM); Integration Reference Point (IRP); Common Object Request Broker Architecture (CORBA) Solution Set (SS)".
[14] 3GPP TS 28.631: "Telecommunication management; Inventory Management (IM) Network Resource Model (NRM) Integration Reference Point (IRP); Requirements".
[15] 3GPP TS 28.632: "Telecommunication management; Inventory Management (IM) Network Resource Model (NRM) Integration Reference Point (IRP); Information Service (IS)".
[16] 3GPP TS 28.633: "Telecommunication management; Inventory Management (IM) Network Resource Model (NRM) Integration Reference Point (IRP); Solution Set (SS) definitions".
[17] SP-250863: Study on SBMA enhancement phase 4
[18] 3GPP TS 32.158: " Design rules for REpresentational State Transfer (REST) Solution Sets (SS)"
[19] RFC 6241 Network Configuration Protocol (NETCONF)
[20] 3GPP TS 32.101: "Telecommunication management;Principles and high level requirements".
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 3 Definitions of terms, symbols and abbreviations
| |
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 3.1 Terms
| For the purposes of the present document, the terms given in TR 21.905 [1] and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in TR 21.905 [1].
example: text used to clarify abstract rules by applying them literally.
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 3.2 Symbols
| For the purposes of the present document, the following symbols apply:
<symbol> <Explanation>
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 3.3 Abbreviations
| For the purposes of the present document, the abbreviations given in TR 21.905 [1] and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in TR 21.905 [1].
<ABBREVIATION> <Expansion>
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 4 Concepts and background
| |
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 4.1 Introduction
| SBMA adopts a modular, service-oriented approach that enables flexible, scalable, and interoperable management solutions.
The architecture for the present study builds upon the SBMA framework in TS 28.533 [2]. All potential enhancements and solutions are expected to comply with the principles outlined in TS 28.533 [2], including service abstraction, interface standardization, and functional decoupling.
The present study aims to preserve architectural continuity with the SBMA baseline model while expanding its applicability to address emerging requirements in 5G Advanced networks.
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 4.2 Management data streaming based on Message Bus technologies
| 3GPP TS 28.532 [3] defines Stage 2 and Stage 3 of the streaming data reporting service. Currently, the 3GPP management system supports WebSocket-based data streaming for PM, tracing, and analytics, establishing a point-to-point transport connection between the streaming data reporting MnS consumer and MnS producer. The producer initiates the connection with the consumer and exchanges metadata (informing the consumer about its identity and the nature of the data to be reported, i.e., streamInfo). After this phase, the producer reports streaming data to the consumer via the established connection.
Using point-to-point WebSocket communication protocol, which operates over a single TCP connection between a client and a server, a separate connection would be established between each consumer and the producer, causing the producer to generate and transmit multiple copies of the same data. The WebSocket protocol is standardized by the IETF in RFC 6455 [8]. It defines WebSocket as a protocol that enables ongoing, full-duplex, bidirectional communication between web servers and web clients over an underlying TCP connection, see [9]. Though a server can maintain multiple WebSocket connections simultaneously and broadcast messages to all connected clients. However, this is not a protocol-level feature, it is an application-level point-to-multipoint behaviour.
Editor's note: It is FFS to investigates the integration and compatibility with the message bus industry solutions by providing a general scope which covers management data types defined in 3GPP TS 28.537 [4] (e.g. performance data, trace data, etc.). This does not imply that the message bus investigated in the present document are intended to replace the currently used WebSocket.
NOTE: The management data includes performance measurements or KPIs defined in TS 28.552 [5] and TS 28.554 [6], trace reporting data defined in TS 32.423 [7] and analytic reporting data defined in TS 28.532 [3].
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 4.3 Message Bus principles integrated in SBMA
| Message Bus principles
A Message Bus is a middleware architecture that enables different systems or services to communicate through a shared messaging infrastructure in a decoupled way. Key principles include the following:
• Loose Coupling: Services interact without knowing each other’s internal details. Enhances modularity and allows independent evolution of components.
• Asynchronous Messaging: Messages are sent without waiting for immediate responses. Improves system responsiveness and fault tolerance.
• Publish/Subscribe Model: Producers publish messages to a topic; consumers subscribe to topics of interest. Enables one-to-many and many-to-many communication.
• Scalability and Resilience: Supports high-throughput, fault-tolerant communication with buffering, retries, and failover etc. Maintains service continuity under load or failure.
Table 4.3-1 shows how the message bus principles are supported in the SBMA to modernize 3GPP network management.
Table 4.3-1: Message Bus principles supported in the SBMA
Message Bus principle
SBMA implementation
Loose Coupling
MnS producers and consumers interact via service exposure. Scaling of one side, e.g. MnS producer, does not affect the other side, e.g. MnS consumer. Adding more number of MnS consumers, does not affect the MnS producer. Error handling and failure recover does not require coupled proprietary implementation between MnS producer and MnS consumer.
Asynchronous Messaging
SBMA supports subscription/notification and streaming interfaces for non-blocking data flow (i.e., MnS producer sends the message data and continues its work without delay, while MnS consumer processes the message independently).
Publish/Subscribe
MnS consumers can subscribe to events or data streams exposed by MnS producers.
Scalability and Resilience
SBMA supports multipoint-to-multipoint streaming, improving efficiency and fault tolerance
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5 Use cases and potential solutions
| |
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5.1 Use case #1: Integration of SBMA with 5GC and 5G Access Network architecture
| |
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5.1.1 Description
| In a 5G Advanced network deployment, a mobile operator adopts an SBMA that leverages a modular reference model as defined in annex A.11 of TS 28.533 [YY]. This model abstracts Management Functions (MnFs) as logical entities, independent of their physical implementation, and enables them to interact exclusively through Management Services (MnSs).
Each MnF registers its MnSs with a service registry, enabling dynamic discovery and selection of services across domains and vendors.
For example:
• A fault management function in the RAN management domain autonomously detects faults using AI/ML models and triggers a healing workflow.
• This MnF can then orchestrate actions across both RAN and Core domains, invoking MnSs such as performance analytics, configuration updates, and resource scaling.
• The orchestration logic is implementation-agnostic, which means it does not rely on where or how the functions are deployed, only on the standardized MnS interfaces.
• The SBMA framework ensures that each MnF clearly declares its role as a producer or consumer of MnSs, supporting modularity and reusability.
In Release-19, a modular reference model has been defined in annex A.11 of TS 28.533 [YY]. Continue evolving the architecture reference model for management and orchestration, to support new requirements such as AI/ML integration, autonomy, and cross-domain orchestration.
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5.1.2 Potential requirements
| REQ-SBMA-ARM-1: SBMA should provide a consolidated view of management architecture.
|
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5.1.3 Potential solutions
| |
a783bfffabe0a6b1e88e24a2cc62e556 | 28.884 | 5.1.4 Evaluation of potential solutions
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.