PoC #3 Demo: 6G-Cloud Network Digital Twins Demo (Lead: NCS) Participants: NCS

[Audio] Hello everyone! In this video, we present the Network Digital Twins demo proof-of-concept demonstration for the 6G-Cloud project. Our demo highlights the methodology of the network digital twin creation and its coexistence in the 6G-Cloud Architectural implementation. The architecture, workflows, and results we present are the outcome of efforts by National Centre for Scientific Reseach Demokritos, as part of ongoing work in Work Packages 2,3 and 5..

[Audio] This presentation will cover: Firstly, the Proof of Concept Introduction and Objectives Secondly, we will present the Network Digital Twin Architectural design. Then we will describe the Demo Workflow. Following that, we will present the initial results from the Physical Network and its Network Digital Twin counterpart. Finally we will analyse the results and discuss the future work..

[Audio] In this part of the presentation we will introduce the POC Introduction and Objectives.

[Audio] Concept: One of the primary objectives of 6G-Cloud is to establish the Network Digital Twin (NDT) concept, a methodology to create virtual models that interact with real networks in real-time, enabling data-driven, closed-loop network automation beyond traditional network simulation. Based on the NDT an AI/ML framework will be built that serves as a comprehensive cognitive plane for the 6G-Cloud architecture. The demo has three main objectives: Firstly to present the NDT generation methodology and Create the NDT of the Physιcal Network testbed. Secondly to evaluate the modelling of the network topology with respective nodes, links, and components. Thirdly to evaluate the integration with the physical network to exchange real-time and historical data for extended data analysis..

[Audio] In the proposed 6G-Cloud Architecture the Network Digital Twin is a building block within the AIMLF framework. The Νetwork Digital twin is implemented as a representation of the physical network related to an NS to accelerate the AI/ML training and optimize the performance of the physical network..

[Audio] In this part of the presentation we will introduce the Network Digital Twin design methodology.

[Audio] The Network Digital Twin (NDT) is an advanced, AI-driven virtual representation of a Physical Network (PN) that continuously synchronizes with real-time data to simulate, analyze, and optimize network operations. This concept is a key enabler for 6G's self-optimizing, predictive, and resilient communication infrastructure. Data Repository : handles all data-related processes necessary for model training and validation. It consists of: Data Collection to gather data from the Physical Network and other sources, Data Preprocessing to transform and clean data for use, Data Management to ensure efficient data retrieval and organization, Data Storage as the central repository for all data within the system. NDT Engine is the core of the NDT module (network analysis, emulation, diagnosis, prediction, simulation, etc.). It Operates through multiple instances (NDT 1, NDT 2, ..., NDT n) that represent different layers or segments of the network. It consists of: Network Modeling Service to simulate network behavior, enabling virtual testing and validation of AI/ML models before they are deployed in real-world settings. Physical Network Mapping Service to serve as an interface between the NDT and the physical network, helping the NDT reflect the physical network's structure and state accurately in simulations. NDT Management (NDT Mgmt) provides tools to monitor, manage, and maintain the NDT's operations. It consist of: The NDT Orchestrator that manages workflows and synchronizes operations across the NDT Engine and Data Repository. The NDT Lifecycle Management that oversees the lifecycle of NDT instances, handling their creation, maintenance, and termination to ensure smooth, uninterrupted operations. The NDT Performance Evaluation that assesses model accuracy and ensures that simulations meet desired performance standards. The NDT Visualization Service for graphical representations of the NDT's data and processes..

[Audio] The NDT Building Blocks are The Data Repository implemented by Prometheus Monitoring System and Time Series Database The NDT Engine consists of the Radio Access Network NDT that is implemented by UERANSIM User Equipment and gNodeB simulator and the core network NDT simulator that is implemented by the Open5GS open source Core Network implementation The NDT Orchestrator is implemented by the NCSR Demokritos developed KATANA open-source Manager and Network Orchestrator. The NDT Visualization Service is implemented by the Grafana visualization and observability solution..

[Audio] The following data will be used to implement and evaluate the NDT creation methodology. Furthermore these data will be used as data flows for the AIMLF framework to enable intelligent automation and real-time system optimization. From the Radio Access Network UE logs GNB Logs Traffic logs NAS messages GTP-U tunnel logs NGAP messages From the Core Network AMF logs SMF logs UPF logs NRF logs PCF logs NSSF logs.

[Audio] Furthermore NDT Data flows include data from Compute Resource for: CPU Utilization Memory Utilization Virtula Machine Status Contained Performance Metrics Energy Consumption End to End Network Service Data Bandwidth Utilization Latency Packet Loss Jitter Network Congestion.

[Audio] In this part of the presentation we will introduce the Demo workflow.

[Audio] Our Demo Set-Up consists of Physical Network Amarisoft Classic (RAN & Core) Two 5G UEs (Huawei P40) Networking Digital Twin Open5GS (Core) UERANSIM (Virtualized UEs, gNBs) Prometheus Database Grafana Visualization and Observability NDT Orchestration Katana Traffic Generators IXIA IxChariot Network Performance Testing (multiple traffic profiles) Computing Infrastructure Computing virtualization cluster Proxmox VE cluster, supported by NCSRD NOC.

[Audio] The workflow for the demo consists of 5 Steps. Going step by step In the first step we deploy the physical network. In the second stop we use the NDT orchestrator to deploy the physical network's network digital twin. In the third step we run similar traffic scenarios in both networks. In step 4 we gather the real time telemetry data from the physical network and the network digital twin. Finally in step 5 we evaluate the NDT performance in relation to the physical network..

[Audio] The first phase of the deployment process involves the initiation of the physical network, during which Physical Network Services (PNS) are activated, and the User Equipment (UE) attachment procedure is carried out. This step is crucial in establishing the fundamental connectivity between the UE and the 5G Core Network (5GC), enabling secure access to network services and mobility management. The process consists of several key sub-steps that ensure successful registration, authentication, and signaling exchange within the network. One of the primary components of this phase is the Non-Access Stratum (NAS) Registration. In this process, the UE establishes a connection with the 5G Core by communicating with the Access and Mobility Management Function (AMF) via the gNB (gNodeB, the 5G base station). Following registration, the Non-Access Stratum (NAS) Authentication takes place to ensure that the UE is legitimate and authorized to access the network. This step is critical in preventing unauthorized access and securing network resources. Additionally, the exchange of NGAP (Next Generation Application Protocol) messages plays a vital role in completing the UE registration process. NGAP facilitates communication between the gNB and the AMF, enabling session management, mobility handling, and signaling necessary for UE attachment. By successfully completing NAS Registration, NAS Authentication, and NGAP Message Exchange, the physical network is fully initialized, and the UE attachment procedure is completed. This sets the stage for further network functions, including PDU session establishment, mobility management, and service provisioning, allowing the UE to fully interact with the 5G network infrastructure..

[Audio] In Step 2, the Network Digital TWIN deployment process is executed using the KATANA Management and Orchestrator, a key component for automating and streamlining network deployments. This phase begins with the upload of the Open5GS Network Service Descriptor (NSD) and Kubernetes Network Function (KNF) to Open Source Mano (OSM), a widely used open-source management and orchestration framework for network function virtualization (NFV). This ensures that the necessary configurations and service descriptions are available for deployment within the 5G core network. Following this, Katana is utilized to upload the cluster credentials, enabling secure communication between the orchestrator and the infrastructure. Once the credentials are in place, the cluster itself is registered with Katana, allowing it to be recognized as a deployment target within the orchestration framework. In parallel, OSM is also registered with Katana, creating a seamless integration between the network service orchestrator and the management platform. With the foundational configurations in place, the final step involves deploying the Open5GS NSD/KNF using Katana, ensuring that the correct service instances and network functions are instantiated within the Kubernetes environment. This deployment requires the provision of a JSON configuration file, which defines the necessary parameters for the service, such as resource allocation, connectivity settings, and operational policies. By leveraging this structured approach, the entire process is automated and optimized, ensuring a scalable, efficient, and reproducible 5G core deployment..

[Audio] During the Network Digital Twin initiation and session establishment phase, we observe a structured and sequential exchange of signaling messages between different components of the 5G network, ensuring seamless connectivity and session establishment. This phase is critical in validating the interaction between the Radio Access Network (RAN) and the 5G Core (5GC), as well as verifying the proper handling of UE session requests. The successful completion of this phase confirms that the network infrastructure is correctly configured and that the UE can establish a data session with the core network. The first category of messages observed in this phase pertains to AMF Connectivity Messages, specifically the gNodeB (gNB) registration messages. These messages are part of the NGAP (Next Generation Application Protocol) setup process, where the gNB establishes a connection with the Access and Mobility Management Function (AMF). The NG Setup Request is sent from the gNB to the AMF, initiating the registration of the RAN node within the 5G Core. The AMF responds with an NG Setup Response, confirming successful registration and allowing the gNB to begin handling UE connections and mobility management procedures. Following the successful registration of the gNB, we observe the exchange of RAN Network NGAP and RRC (Radio Resource Control) messages, particularly the NG Setup Registration messages. These messages are crucial for establishing control-plane communication between the gNB and the AMF, enabling proper coordination of UE signaling and resource management. The NGAP Initial UE Message is exchanged when a UE attempts to connect to the network, initiating NAS signaling. Additionally, the RRC Setup procedure ensures that the UE can communicate with the gNB, allowing further interactions with the core network. Once the network connectivity is established at both the RAN and Core levels, we observe UE Session Establishment messages, particularly PDU Session Setup messages. This process is essential for enabling data transfer and internet access for the UE. When the UE requests a PDU session, the Session Management Function (SMF) interacts with the AMF and the User Plane Function (UPF) to allocate the necessary resources. The PDU Session Resource Setup Request is sent from the AMF to the gNB, instructing it to configure the UE's data path. Upon successful allocation of network resources, the gNB responds with a PDU Session Resource Setup Response, allowing the UE to initiate data exchange through the 5G network..

[Audio] In this part of the presentation we will discuss the initial results of the demo..

[Audio] Our testing scenario is designed to evaluate the performance, reliability, and accuracy of the Network Digital Twin (NDT) by comparing its behavior to that of the Physical Network under identical conditions. This setup involves the use of two User Equipment (UE) nodes, each of which operates within both the Physical Network and its corresponding NDT environment. By running the same traffic patterns in both networks, we aim to assess the fidelity of the NDT in replicating real-world network conditions and its ability to provide actionable insights for network optimization. Each UE node is assigned a distinct traffic mix to simulate diverse user behaviors and network usage patterns. This variation in traffic ensures a more comprehensive evaluation of the network's performance under different conditions. Additionally, each UE is provisioned with a specific Quality of Service (QoS) Service Level Agreement (SLA), ensuring that network resources are allocated according to predefined throughput requirements. Specifically: UE 1 is assigned a QoS SLA of 17 Mbps, ensuring higher bandwidth availability for applications requiring high-speed data transfer. UE 2 is allocated a QoS SLA of 10 Mbps, simulating a use case with moderate bandwidth demands. By implementing these differentiated QoS profiles, we can assess how effectively both the Physical Network and the NDT manage resource allocation, maintain service quality, and handle traffic prioritization in varying network conditions..

[Audio] Physical Network Testing and Traffic Flow Analysis The Physical Network Testing scenario is visually presented in the video, where the traffic flow analysis is conducted using the Ixia Chariot network performance testing tool and the Amarisoft physical network web interface. This setup provides a comprehensive view of uplink data traffic behavior, allowing for precise monitoring and performance evaluation of the 5G network infrastructure. Traffic Flow Analysis with IXIA Chariot In the top half of the video, the traffic flow scenario is displayed, illustrating how the Ixia Chariot network testing tool is used to generate and measure uplink TCP traffic from each User Equipment (UE) node towards the IXIA Chariot server. This configuration represents a real-world network environment, where data is transmitted from the UEs to the core network, enabling the analysis of throughput, latency, and overall network performance. The IXIA Chariot tool plays a crucial role in validating the network's ability to handle concurrent data flows from multiple users while maintaining the expected Quality of Service (QoS) levels. Traffic Monitoring via the Amarisoft Physical Network Interface In the bottom half of the video, the Amarisoft physical network web interface is displayed, providing real-time visualization of uplink traffic flows as measured from the gNodeB (gNB). This interface captures and presents network traffic metrics, allowing for detailed observation of the data being transmitted. As the scenario unfolds, we can see the real-time changes in uplink data traffic flows as generated by the two UEs. Uplink Throughput Representation The uplink traffic throughput for each UE is distinctly represented using color-coded indicators: UE1's uplink traffic throughput is represented by the red line, which maintains a throughput of 17 Mbps. UE2's uplink traffic throughput is depicted by the green line, which shows a throughput of 10 Mbps. The total uplink throughput, which is the summation of UE1 and UE2 traffic, is represented by the blue line, indicating an aggregate throughput of 27 Mbps. This real-time visualization provides valuable insights into how the network handles multiple concurrent uplink traffic streams and verifies that the Physical Network is functioning as expected, maintaining the desired QoS levels for each UE. Through this testing, we can confirm that the network successfully distributes resources while sustaining stable and predictable data flows..

[Audio] Network Digital Twin (NDT) Performance Evaluation In this phase, we replicate the exact same testing scenarios within the Network Digital Twin (NDT) environment to compare its performance against the Physical Network. The Grafana dashboard is used to visualize real-time network performance metrics, which are collected from the Prometheus Database live logs. This allows for a comprehensive analysis of traffic flow, system resource utilization, and overall network behavior within the NDT framework. Traffic Flow Analysis in the NDT Similar to the Physical Network test, the TCP uplink traffic is generated by each User Equipment (UE) and directed towards a server interconnected with the NDT. This setup mirrors the real-world data flow scenario and enables us to monitor network performance at different layers, ensuring that the NDT replicates the physical network with high fidelity. Grafana Dashboard Overview The Grafana dashboard is divided into two sections: 🔹 Left Side of the Dashboard – Uplink Traffic and System Resource Utilization The left section of the Grafana dashboard provides insights into individual UE performance, total throughput, and system resource utilization. Key dashboards include: UE1/UE2 Dashboard: Displays the uplink throughput for each User Equipment (UE) separately, allowing for real-time tracking of their data transmission rates. Ogstun Dashboard: Measures the total throughput at the User Plane Function (UPF) interface, ensuring that the network's data-handling capacity aligns with expectations. Memory Usage Dashboard: Tracks the memory consumption of the Virtual Machine (VM) running the NDT, helping to assess the system's resource efficiency. From these metrics, we conclude that the NDT achieves a performance level similar to that of the Physical Network, validating its capability to accurately emulate real-world conditions. 🔹 Right Side of the Dashboard – Network Traffic and System Performance Metrics The right section of the Grafana dashboard focuses on system-wide network behavior and processing capacity. Key dashboards include: Network Traffic Dashboard: Monitors the throughput metrics of the Virtual Machine (VM) running the NDT, ensuring data flows are consistent with the expected values. CPU Usage Dashboard: Tracks the CPU utilization of the NDT VM, providing insights into processing efficiency and computational overhead. Packet Loss Dashboard: Measures the packet loss at the UPF interface, a critical metric for determining network reliability and quality of service. Detailed Traffic Flow Analysis By focusing on the UE1/UE2 Dashboard and Ogstun Dashboard, we observe the uplink data traffic flow as the scenario evolves. UE1 uplink traffic throughput is represented by the green line, maintaining a steady rate of 17 Mbps. UE2 uplink traffic throughput is represented by the yellow line, consistently reaching 10 Mbps. Total throughput, as shown in the Ogstun Dashboard, is the sum of UE1 and UE2 traffic, reaching a combined rate of 27 Mbps, aligning with the values observed in the Physical Network test. Final Validation and Conclusion The successful replication of traffic patterns and performance metrics between the Network Digital Twin (NDT) and the Physical Network confirms that the NDT operates with a high degree of accuracy. By analyzing the throughput, system resource utilization, and packet loss metrics, we validate that the NDT provides a reliable and scalable environment for network performance evaluation. This ensures that future tests and optimizations can be conducted within the Digital Twin framework with confidence..

[Audio] Analysing the results we conclude that NDT network is successfully deployed. NDT produces similar performance results with physical network. Expandable NDT baseline for further implementation. For the continuation of the project our future plans are : Extended implementation of the proposed methodology Multiple UEs Different traffic mixes Diverse network topologies involving various Network Services / Network Slices Ability to support the training, evaluation, and validation of AI/ML models for data collection, federated training, and inference for optimized network observability. Support of an AIML framework to incorporate functions such as model training, performance monitoring, AI/ML models orchestration facilitating easy deployment. Continuous integration of physical network and NDT with AIML framework for historical and real-time data exchange to support AIML-driven insights..