The correct configurations of optical infrastructure are the cornerstone for telecoms to stably provision large-capacity and long-distance service pipelines. Although traditional configuration workflows can be automated using predefined message templates and scripting, the introduction of new multi-vendor devices or required updates to network configurations forces administrators to manually review extensive documentation to implement the necessary procedures. Fundamentally, these processes are a form of natural language understanding and transformation. To address the high effort-cost, time-investment, and cross-vendor interoperability challenges, we introduce large language models (LLMs) with additional reasoning enhancement to achieve Automation-from-Intentions . Specifically, we design a reinforcement-learning-based fine-tuning paradigm of an LLM and evaluate its performance on a deployed cross-vendor optical network. The results show perfect configuration accuracy—achieving a score of 100 without hallucinations—across a 12-vendor, 256-node network. By reducing reliance on vendor-provided professional services and lowering manual operational effort, our proposal can effectively decrease both CAPEX and OPEX for telecom operators.
Rate-flexible coherent passive optical networks (PONs) introduce heterogeneous physical-layer conditions and dynamic rate configurations across optical network unit (ONU) branches. These variations require robust joint monitoring of entropy and SNR. To address this issue, we propose an adaptive weight learning (AWL) method based on a convolutional neural network (CNN) for multi-parameter learning. Joint monitoring of entropy and SNR in a probabilistic constellation shaping (PCS)-based coherent transmission network is performed. A point-to-multipoint (PtMP) access system with a baud rate of 32 Gbaud and 40 km coverage is experimentally demonstrated over back-to-back, 26.7 km, and 40 km links, achieving data rates from 134.4 to 166.4 Gbps at entropies ranging from 4.2 to 5.2 bits. Dual-regression analysis in multi-parameter monitoring yielded lower mean absolute errors, reducing the entropy error by 0.028 bits and SNR error by 0.111 dB compared to single-parameter learning. Additionally, AWL demonstrated robustness and effectiveness across various scenarios, including mixed-modulation formats, extreme SNR conditions, chromatic dispersion, laser phase noise, transceiver nonlinearity, and supplementary upstream-oriented scenarios. Finally, the monitoring parameters are scalable to accommodate future optical access network demands.
We propose an integrated sensing and communication (ISAC) architecture capable of simultaneously performing distributed acoustic sensing (DAS) and distributed temperature sensing (DTS) in a single-mode fiber (SMF) and experimentally evaluate its coexistence feasibility using a discrete-component proof-of-concept setup. This system supports parallel multi-parameter sensing based on phase-sensitive optical time-domain reflectometry ($\phi$-OTDR) and Raman optical time-domain reflectometry (ROTDR). The present experiment is conducted using discrete commercial components, while integrated polarization–wavelength de/multiplexing is retained only as a possible future compact implementation route. Hybrid amplification techniques, including Raman distributed amplification and optical-bias-assisted superimposed enhancement, are introduced to improve the communication link budget and ROTDR detectability under a reduced sensing-probe power. In addition, polarization-diversity reception is analyzed as a potential extension to mitigate polarization fading and improve the stability of Rayleigh-based vibration sensing. Experimental results demonstrate system-level coexistence of $\phi$-OTDR-based DAS, ROTDR-based DTS, and single-polarization 2.5 Gbit/s non-return-to-zero/intensity-modulation direct-detection (NRZ/IMDD) communication over a 10 km SMF link. The simplified proof-of-concept experiment achieves a spatial resolution of 5 m, a temperature uncertainty of 2.8°C, and a vibration-sensing signal-to-noise ratio (SNR) of 12.1 dB under a sensing-probe peak power of 100 mW, while no bit errors are observed in the communication branch during the experimental test. Fiber-optic communication and sensing integration technology can support traditional applications such as oil and gas pipelines, power lines, and environmental monitoring, and may provide useful sensing capability for emerging scenarios such as digital twins and intelligent optical networks.
This feature issue of the Journal of the Optical Networking presents articles that span three complementary areas: open datasets and models, open software and benchmarking platforms, and benchmarking studies and reference methodologies.
This publisher’s note contains a correction to J. Opt. Commun. Netw. 18 , B108 ( 2026 ) 10.1364/JOCN.585552 .
Multi-band space division multiplexed (MB-SDM) elastic optical networks provide a practical path to ultra-high-capacity backbone transport by combining spectral expansion beyond the C-band with spatial multiplexing over weakly coupled multi-core fibers (MCFs). Yet, the benefits of these two scaling dimensions are strongly shaped by physical layer impairments and cannot be assessed independently. This paper presents a quality of transmission (QoT)-aware planning framework for MB-SDM backbone networks that jointly evaluates C, $\text{C} +\text{L}$, $\text{C} +\text{S}$, and $\text{C} +\text{L} +\text{S}$ transmission over 4-, 7-, 13-, and 19-core MCFs. The framework relies on a unified generalized signal-to-noise ratio model including amplified spontaneous emission, nonlinear interference, inter-channel stimulated Raman scattering, inter-core crosstalk (ICXT), transceiver limitations, and filtering and aging penalties. A QoT-aware spectral–spatial resource allocation algorithm is evaluated under dynamic traffic, where modulation formats are precomputed based on precise impairment modeling with launch power optimization, and bandwidth slicing is applied; the ICXT model is further validated through experiments on a commercial 19-core MCF. Results on the U.S. backbone topology show that the preferred band configuration depends strongly on the core count scenario. For 4- and 7-core MCFs, $\text{C} +\text{L}$ provides better throughput and lower blocking than $\text{C} +\text{S}$. For 13-core MCFs, $\text{C} +\text{S}$ becomes more effective because the L-band is more strongly affected by ICXT. For 19-core MCFs, the system becomes ICXT-dominated, causing severe degradation in C and $\text{C} +\text{L}$, while $\text{C} +\text{S}$ achieves the best overall performance. Notably, despite the larger spatial resource pool, the maximum throughput in the $\text{C} +\text{L} +\text{S}$ scenario decreases from 107.94 Pbps in the 13-core case to 44.76 Pbps in the 19-core case, confirming that aggressive spatial densification can reduce usable network capacity.
Motivated by the recent trend toward capacity scaling through spatial-division multiplexing (SDM) and band-division multiplexing (BDM), we propose a cost-efficient SDM/BDM network planning method based on hierarchical optical cross-connects (HOXCs). Each HOXC node consists of a wavelength cross-connect (WXC) and a band cross-connect (BXC), where additional band-granular switching reduces the number of required wavelength-selective switches (WSSs). Based on the node architecture and incorporating a band-dependent component cost model, the wavelength-/band-paths are configured so that the total network cost is minimized. We also developed a heuristic algorithm that uses auxiliary graphs to accurately model the considered system architecture and component costs, achieving an average optimality gap of less than 8.8%. Numerical simulations demonstrated that the proposed method reduces network costs by up to 20% while reducing the number of WSSs by up to 88.2% in the considered topologies. We also found that the effectiveness of the HOXCs and the number of multiplexed bands differ depending on traffic volume, average link length, and cost parameters.
The explosive development in artificial intelligence (AI) like large language models (LLMs) requires massive graphics processing units (GPUs) during training, making distributed machine learning (DML) across multiple data centers (DCs) a practical approach to obtaining sufficient training resources. Among cross-DC DML techniques, pipeline parallelism (PP) is a natural solution, which partitions LLM layers across GPUs and orchestrates point-to-point communication between adjacent nodes. However, the high communication latency induces severe GPU idle time (“bubble”) due to poor communication and computation time overlapping in inter-DC PP. While prior works have explored partitioning and reordering computation tasks, the resource allocation for inter-DC communication between pipelined computing tasks over optical networks has been overlooked, resulting in higher GPU idle time and request congestion. Therefore, it is extremely important and challenging to propose an advanced inter-DC network resource allocation solution under the complex PP workload. In this work, we propose a cross-domain communication-bound-aware (CD-CBA) PP scheduling framework for multi-DC optical networks. Simulation results show that the proposed framework lowers the LLM training iteration time by up to 31.25%, decreases the bubble ratio by up to 11.81%, and reduces the transmission request blocking by up to 19.33%.
Efficient resource allocation via routing, modulation, and spectrum allocation (RMSA) is critical for elastic optical networks (EONs). While analytical relative cost models have shown promise for adaptive RMSA, they often suffer from high computational complexity and restrictive assumptions that limit practical applicability. In this study, we present a neural-network (NN)-based approach that learns the relative cost function directly from data. The proposed method eliminates the need for iterative optimization and specialized teletraffic modeling, offering a more flexible and accurate estimate of the trade-off between immediate and future network revenue. We describe the systematic generation of a large-scale simulation dataset, the design of a feature set capturing link states and spectral fragmentation, and the training of a robust NN model. Extensive evaluations on the NSFNET and COST239 topologies demonstrate that the proposed neural-network-based RMSA algorithm reduces the blocked bandwidth ratio (BBR) by an average of $34\%$ while improving fairness by $70\%$ compared to established benchmarks. The model’s adaptability and offline training paradigm highlight its potential for deployment in operational networks.
Optical fiber sensing on deployed cables is gaining significant interest as a means to enhance network integrity and expand environmental monitoring capabilities. To fully achieve these goals, sensing signals must coexist with data transmission. While phase and state-of-polarization analysis are promising candidates for this integration, their practical exploitation presents several challenges. Because these methods measure the integrated deformation along the entire cable, sensitivity is often limited by high environmental noise floors. Furthermore, event interpretation and localization require more sophisticated analytics than conventional distributed fiber sensing. Incorporating fiber sensing into standard network architectures also necessitates a novel telemetry management system capable of handling heterogeneous datasets, high data rates, and stringent synchronization requirements. We present progress on these aspects, based on multi-year analysis of fiber-sensing data collected in field environments. Our results confirm that integrated sensing achieves the sensitivity required to detect mid- and low-magnitude seismic events within a range of tens of kilometers from the fiber, holding the potential to support seismic monitoring services in terrestrial areas. Additionally, we demonstrate the detection and localization of anomalous vibrations and anthropogenic events and the development of an autonomous anomaly-detection tool suitable for real-time alerting. Finally, we outline the requirements for an advanced network management system designed to accommodate sensing alongside conventional telemetry, supporting the transition from proof-of-concept experiments to robust, large-scale applications.
TopoLib is an open-source Python library providing a curated collection of core network topologies and analytical traffic models for rapid network evaluation. The library includes 60 optical core networks sourced from operator websites, scientific literature, and research projects. Each topology contains node geo-locations, population figures, and per-node counts of data centers (DCs) and Internet exchange points (IXPs), along with link interconnections and distances. Three analytical traffic models are provided: two state-of-the-art models [gravitational and multi-period traffic (MPT)], and the proposed region aggregation model (RAM), which estimates traffic based on population, DCs, and IXPs. Applied to a topology, these models produce a traffic matrix suitable for network evaluation and cross-study comparison. All data can be exported for external use, and researchers may load their own topologies and modify model parameters to suit their specific needs. To demonstrate the library, three topologies are evaluated across nine distinct cases, covering three analyses: a comparison of traffic generated by each model, a sensitivity analysis of the RAM components, and a case study integrating TopoLib with MONET, a multi-period optical network planning tool based on Flex Net Sim [J. Opt. Commun. Netw.18, D16 (2026)], showcasing how the library can support exhaustive network utilization and infrastructure deployment studies.
As the demand for high-reliability and high-throughput optical transport networks intensifies with the rise of artificial intelligence (AI) and large-scale distributed applications, conventional wavelength-switched optical network (WSON) protection schemes, which rely on slower restoration times, become inadequate. In this work, we propose a novel, to our knowledge, sub-50-ms WSON protection scheme that integrates fast protocol processing, high-speed wavelength selective switching (WSS), and ultra-fast optical digital signal processing (DSP) reconstruction. By pre-configuring protection paths and parallelizing protocol processing, the proposed scheme eliminates key latency bottlenecks, ensuring rapid, real-time protection switching. We demonstrate its performance through both laboratory and field trials, achieving end-to-end protection switching within 50 ms, thus meeting the stringent latency requirements of AI applications. This work introduces a scalable and efficient solution for next-generation optical networks, addressing the challenges of real-time, fault-tolerant communication in a rapidly evolving digital landscape.
