Optical Network Digital Twin -- Practical Use Cases and Architecture

arXiv:2511.06368 · eess.SY, cs.SY · Submitted 2025-11-09 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Optical Network Digital Twin -- Practical Use Cases and Architecture".

Dev: Here's the gist of the paper. Optical Network Digital Twin – Practical Use Cases and Architecture discusses an evolving Optical Network Digital Twin architecture that enables dynamic and Quality of Transmission aware operation beyond conventional management The proposed ONDT architecture integrates QoT monitoring and historical data within a unified digital representation of the network, maintaining consistency between the physical and digital domains Evolving Operational Requirements The paper presents a three-step evolution of ON operation toward the AI era and discusses how ONDT-based operation enables new operational capabilities at each stage These stages include: 1.

Rosa: Operator-driven optimization, which enables cost reduction and flexible service-level configuration through freedom of device selection and minimization of unnecessary margins, supported by QoT visualization and real-time performance awareness.

Rosa: User–operator collaboration, which supports on-demand lightpath design and optimization between arbitrary locations, including protocols and required operation time based on prior experimental and implementation results.

Paper summary: Rosa: So we're looking at this paper called "Optical Network Digital Twin -- Practical Use Cases and Architecture," which is about how to evolve optical networks for the AI era Dev it talks about a new digital twin approach that allows for dynamic operation and quality of transmission aware management beyond what we see now Rosa

Dev: Right, it claims this architecture integrates quality of transmission monitoring and historical data into one unified digital representation of the network, keeping the physical network consistent with the digital one Dev

Rosa: The paper outlines three steps in how optical network operation is evolving toward the AI era Rosa first you have operator-driven optimization which lets them reduce costs and set service levels flexibly by picking devices and cutting unnecessary margins, supported by quality of transmission visualization and real time performance awareness Rosa

Dev: Then there's user operator collaboration where you can design and optimize lightpaths on demand between any locations, including protocols and required operation times based on what has already been done experimentally or implemented Dev

Rosa: And finally, there's operator operator collaboration which lets different operators connect across domains through coordinated operation between their controllers, emphasizing the importance of efficiently using the optical spectrum across those different operators Rosa

Taro: It sounds like they are moving from just managing a fixed ring network to something much more flexible for things like data center interconnects and connecting multiple service providers Taro

Dev: Exactly. They show how this digital twin model is based on the AWGN assumption, where quality of transmission estimation can be done by evaluating the generalized signal-to-noise ratio, or GSNR, which acts as a characterization metric for optical network impairment Dev

Rosa: They mention that accurate algorithms for estimating quality of transmission can already be implemented in GNPy, accounting for things like the signal power profile, amplified spontaneous emission noise, and nonlinear interference noise along the optical line system components Rosa

Taro: So how does this GSNR estimation work in practice when you're dealing with different technologies? Taro

Dev: The end to end quality of transmission for each transmitter receiver pair over a given link is determined by the received optical power and that GSNR, while things like modulation format, baud rate, and intrinsic noise are treated as fixed characteristics Dev

Rosa: They then show three specific practical use cases for this in metro optical networks Rosa first there's multi generation interoperability in alien wavelength scenarios where you figure out compatibility across different vendors and generations of transmitters receiver pairs based on real time quality of transmission estimation instead of just vendor specific assumptions Rosa

Dev: That lets operators use unused wavelength channels in existing optical line systems and deploy higher rate transmitters even in systems originally designed for lower capacities Dev

Paper summary: Taro: What about the second case, the data center exchange with topological flexibility and on demand provisioning? How quickly can you set up a path there? Taro

Rosa: This use case shows rapid lightpath establishment within about several minutes based on that AWGN model quality of transmission estimation, where the digital twin architecture dynamically picks paths and configures transmitters based on current network conditions for continuous resource optimization Rosa

Dev: The third case is multi operator interconnection across domains, where they achieve consistent end to end quality of transmission estimation across domain boundaries without needing full disclosure of internal network details Dev

Rosa: The architecture includes quality of transmission visualization which turns physical layer performance metrics into something intuitive and actionable for operators, letting them centrally manage the quality of multiple transmitters receiver pairs regardless of their differences in characteristics or the end to end light paths they traverse Rosa

Taro: They also incorporate two types of optical line system models—managed and unmanaged—to help use existing optical infrastructures efficiently Taro

Dev: Looking ahead, they flag some challenges. ones include adding other impairments like power degradation from the WSS in a ROADM, and dealing with linear crosstalk caused by the WSS Dev

Rosa: For unmanaged optical line systems, future work needs to handle scenarios with many wavelengths, variations in gain among channels, and power fluctuations that happen suddenly when a channel shuts down Rosa

Taro: In multi operator environments too, they point out that defining the interfaces and demarcation points will be essential for security and management of shared line system information Taro

Dev: And realizing fully automated digital twin operation will likely involve research on dynamic optical frequency allocation and linking it with upper layer orchestrators Dev

Rosa: So, to wrap up this paper "Optical Network Digital Twin -- Practical Use Cases and Architecture," they propose a GN model-based digital twin architecture that lets you manage quality of transmission flexibly for every end to end light path Rosa

Dev: It supports this step by step evolution from ring networks to data center interconnects and multi operator connections toward next generation optical networks Dev

Rosa: The authors are Nishizawa, Mano, Anazawa, Matsumura, Sasai, Namiki, Briantcev, Ambrosone, Le Rouzic, Melin, Gonzalez-de-Dios Fernandez-Palacios Zhang Akahoshi Grammel D’Amico Borraccini Ruffini Kilper Curri Rosa

Dev: Thinking about the title and the authors it seems they are really focused on making these digital twin concepts practical for real world optical networks rather than just theoretical models Dev

Rosa: The implication is that we can start moving beyond just reliable long haul communication toward a system where quality of transmission is managed dynamically based on what's happening in the network right now Rosa

Conclusion: Rosa: So this paper, "Optical Network Digital Twin -- Practical Use Cases and Architecture," is about taking optical networks and building a digital twin for them, so they can manage things dynamically instead of just reacting to problems after they happen.

Dev: Right, it’s looking at how to make the whole network—the physical stuff—match a digital representation, which lets you test things out in the model first before you mess with the actual hardware.

Rosa: They laid out three big steps for how we evolve optical network operation, starting with operator optimization for cutting costs and setting service levels flexibly.

Dev: Then user operator collaboration where you can design lightpaths on demand between any two points based on what has already been done experimentally.

Rosa: And finally, the third step is when different operators need to connect across their domains, coordinating how they use the optical spectrum together.

Dev: This architecture uses quality of transmission monitoring and historical data all in one digital twin so it keeps the physical network consistent with that digital view.

Rosa: The authors are showing concrete examples like testing multi-generation interoperability across different vendors without needing all those vendor specific assumptions.

Dev: And they show how you can get a data center exchange set up really fast, within minutes, because the digital twin picks the best path based on current conditions.

Rosa: It’s about building something that supports this step by step evolution toward next generation optical networks.

Dev: So what does this mean for us right now is having a tool that lets you see and adjust the quality of transmission across a whole complex system in one place.

Hideki Nishizawa, Toru Mano, Kazuya Anazawa, Tatsuya Matsumura, Takeo Sasai, Masatoshi Namiki, Dmitrii Briantcev, Renato Ambrosone, Esther Le Rouzic, Stefan Melin, Oscar Gonzalez-de-Dios, Juan Pedro Fernandez-Palacios, Xiaocheng Zhang', 'Keigo Akahoshi', 'Gert Grammel', 'Andrea D’Amico', 'Giacomo Borraccini', 'Marco Ruffini', 'Daniel Kilper',

NTT Laboratories · Trinity College Dublin · Politecnico di Torino · Orange Labs · Telia Company AB · Telefonica Innovacion Digital, Spain · NTT Docomo Business, Inc. · KDDI Research, Japan · Juniper Networks Germany · NEC Laboratories America USA

eess.SY, cs.SY

Submitted: 2025-11-09

Updated: 2026-11-05

Comments: 7 pages, 5 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: Optical Network Digital Twin – Practical Use Cases and Architecture discusses an evolving Optical Network Digital Twin architecture that enables dynamic and Quality of Transmission aware operation…

Key concepts

Optical Network Digital Twin (ONDT)
A digital model that mirrors the physical optical network. It combines real-time performance metrics with historical data to provide a consistent digital representation of the network, allowing for advanced management beyond conventional methods.
Quality of Transmission (QoT) Estimation
A method used to assess the quality of an optical signal transmission. It relies on calculating the Generalized Signal-to-Noise Ratio (GSNR), which accounts for impairments like ASE noise and nonlinear interference, to determine the end-to-end performance of a link.
Stepwise Evolution of ON Operation
A three-stage progression of network management driven by AI. It moves from operator optimization (cost reduction) to user-operator collaboration (on-demand lightpath design) and finally to multi-operator interconnection for coordinated spectrum utilization.

Terminology

Summary

Optical Network Digital Twin – Practical Use Cases and Architecture discusses an evolving Optical Network Digital Twin architecture that enables dynamic and Quality of Transmission aware operation beyond conventional management The proposed ONDT architecture integrates QoT monitoring and historical data within a unified digital representation of the network, maintaining consistency between the physical and digital domains

Evolving Operational Requirements

The paper presents a three-step evolution of ON operation toward the AI era and discusses how ONDT-based operation enables new operational capabilities at each stage These stages include:

  1. Operator-driven optimization, which enables cost reduction and flexible service-level configuration through freedom of device selection and minimization of unnecessary margins, supported by QoT visualization and real-time performance awareness. p. 1

  2. User–operator collaboration, which supports on-demand lightpath design and optimization between arbitrary locations, including protocols and required operation time based on prior experimental and implementation results. p. 1

  3. Operator–operator collaboration, which enables multi-operator interconnection across domains through coordinated operation between operator controllers, while highlighting the importance of efficient utilization of optical spectrum across operators. p. 1

Digital Twin Model and QoT Estimation

The Digital Twin Model is based on the AWGN assumption, where the OLS QoT estimation can be obtained evaluating the generalized signal-to-noise ratio (GSNR) that serves as characterizing metric of the OLS transmission impairment Accurate algorithms for OLS QoT evaluations have been already implemented in GNPy [3], accounting for the main contributions to the GSNR, such as the signal power profile, amplified spontaneous emission (ASE) noise and nonlinear interference (NLI) noise, evaluated along the OLS components. The E2E QoT of each TRx over a given link is determined by the received optical power and the GSNR, while TRx parameters such as modulation format, baud rate, and intrinsic noise are treated as fixed characteristics. p. 4

Practical Use Cases

The paper details three representative practical use cases for ONDT in metro ONs

(a) Multi-Generation Interoperability in Alien Wavelength Scenarios involves determining interoperability across multi-vendor and multi-generation TRxs based on real-time QoT estimation rather than vendor-specific assumptions. This allows operators to exploit unused wavelength channels in existing OLS and deploy higher-rate TRxs even in systems originally designed for lower per-wavelength capacities. p. 2

(b) Data Center Exchange with Topological Flexibility and On-Demand Provisioning enables rapid light path establishment within approximately several minutes based on AWGN-model-based QoT estimation. p. 3 The ONDT-enabled DCX architecture dynamically selects feasible paths and configures TRxs based on current network conditions, supporting continuous optimization of network resources. p. 3

(c) Multi-Operator Interconnection Across Domains enables consistent E2E QoT estimation across domain boundaries without requiring full disclosure of internal network details. p. 3 This capability improves the utilization of optical spectrum and enhances flexibility in service provisioning. p. 3

Architecture Components

The ONDT architecture includes QoT visualization, which transforms physical-layer performance metrics into intuitive and actionable information for operators. This visualization enables operators to centrally manage the quality of multiple TRxs regardless of differences in their characteristics, generations, or the E2E light paths they traverse. p. 4 The architecture also incorporates two types of optical line system (OLS) models—managed and unmanaged—to enable efficient utilization of existing optical infrastructures. p. 1

Future Challenges

Future challenges include incorporating other impairments such as PDL induced by the WSS in a ROADM, and addressing linear crosstalk induced by the WSS. For unmanaged OLSs, future work should address scenarios with numerous wavelengths, gain variations among channels, and power fluctuations caused by sudden channel shutdowns. p. 5 In multi-operator environments, defining interfaces and demarcation points will be essential for ensuring security and management of shared line-system information. p. 5 Research on dynamic optical-frequency allocation and linkage with upper-layer orchestrators is also important to realize fully automated ONDT-based operation. p. 5

Conclusion

ONs, which make up the foundation of communication infrastructure, have emphasized high reliability and long-term operation This work proposed a GN-model-based ONDT architecture that enables flexible QoT management for each E2E light path The proposed approach supports a stepwise evolution—from ring networks to data center interconnects and multi-operator connections—toward next-generation ONs.

Acknowledgments

This work is supported in part by a research grant from Taighde Eireann (Research Ireland) under 22/FFP-A/10598 ´ (Twilights), 18/RI/5721 (Open Ireland), 13/RC/2077 P2 at CONNECT: The Research Ireland Centre for Future Networks, and from the National Institute of Information and Communications Technology (NICT), Japan, under JPJ012368G50201 and JPJ012368C09001.

References

[1] K. Aubrey et al. p. 5 (2025, May) Turbocharge LLM training across long-haul data center networks with NVIDIA NeMo framework. p. 5 Accessed: 2025-10-18

[2] IOWN Global Forum. (2024, Jul. p. 5) Remote media production for broadcast industry use case. p. 5 Accessed: 2025-10-18

[3] V. Curri, “GNPy model of the physical layer for open and disaggregated optical networking,” Journal of Optical Communications and Networking, vol. p. 5 14, no. 6, pp. C92–C104, 2022

[4] T. Iye et al., “Open wireless digital twin: End-to-end 5g mobility emulation in O-RAN framework,” arXiv preprint, Mar. p. 5 2025, arXiv:2503. p. 512177

[5] Ericsson. (2025) Chunghwa telecom and ericsson jointly advance 5g slicing automation for industrial digitalization. p. 5 Accessed: 2025-02-21

[6] R. Valalta et al. p. 5, “Applying digital twins to optical networks with cloudnative SDN controllers,” IEEE Communications Magazine, vol. p. 5 61, no. 12, pp. 128–134, 2023

[7] G. Borraccini et al. p. 0, “Experimental demonstration of partially disaggregated optical network control using the physical layer digital twin,” IEEE Transactions on Network and Service Management, vol. p. 5 20, no. 3, pp. 2343–2355, 2023

[8] K. Kaeval et al. p. 5, “Operation of optical spectrum as a service in disaggregated and multi-operator environments [invited],” Journal of Optical Communications and Networking, vol. p. 5 17, no. 1, pp. A46–A58, 2025

[9] H. Nishizawa et al. p. 0, “Dynamic optical path provisioning for alien access links: Architecture, demonstration, and challenges,” IEEE Communications Magazine, vol. p. 5 61, no. 12, pp. 84–90, 2023

[10] ——, “Fast WDM provisioning with minimal probing: the first field experiments for DC exchanges,” Journal of Optical Communications and Networking, vol. 16, no. 2, pp. 233–242, 2024

[11] ——, “Semi-automatic line-system provisioning with an integrated physical-parameter-aware methodology: field verification and operational feasibility,” Journal of Optical Communications and Networking, vol. 16, no. 9, pp. 894–904, 2024

[12] T. Mano et al. p. 0, “Modeling transceiver BER-OSNR characteristic for QoT estimation in short-reach systems,” in 2023 International Conference on Optical Network Design and Modeling (ONDM). p. 6 IEEE, 2023, pp. p. 5 1–3

[13] T. Sasai et al. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Operator-driven optimization enables cost reduction and flexible service-level configuration through freedom of device selection and minimization of unnecessary margins, allowing operators to adapt transmission performance to application requirements, supported by QoT visualization and real-time performance awareness.

  2. User–operator collaboration supports on-demand lightpath design and optimization between arbitrary locations, including protocols and required operation time based on prior experimental and implementation results.

  3. Operator–operator collaboration enables multioperator interconnection across domains through coordinated operation between operator controllers, while highlighting the importance of efficient utilization of optical spectrum across operators.

  4. The ONDT architecture can continuously reflect the network state and enable closed-loop, QoT-aware control, where control decisions are fed back to and applied in the physical network.

  5. The system can dynamically select feasible paths and configure TRxs based on current network conditions when a new connection request is triggered, moving from per-connection provisioning toward a globally optimized and operator-driven network operation.

  6. ONDT enables an operator-driven approach where interoperability across multi-vendor and multi-generation TRxs is determined based on real-time QoT estimation rather than vendor specific assumptions.

Sources

Related papers