Operation Of Optical Networks Research Articles

When Large Language Models Meet Optical Networks: Paving the Way for Automation

Since the advent of GPT, large language models (LLMs) have brought about revolutionary advancements in all walks of life. As a superior natural language processing (NLP) technology, LLMs have consistently achieved state-of-the-art performance in numerous areas. However, LLMs are considered to be general-purpose models for NLP tasks, which may encounter challenges when applied to complex tasks in specialized fields such as optical networks. In this study, we propose a framework of LLM-empowered optical networks, facilitating intelligent control of the physical layer and efficient interaction with the application layer through an LLM-driven agent (AI-Agent) deployed in the control layer. The AI-Agent can leverage external tools and extract domain knowledge from a comprehensive resource library specifically established for optical networks. This is achieved through user input and well-crafted prompts, enabling the generation of control instructions and result representations for autonomous operation and maintenance in optical networks. To improve LLM’s capability in professional fields and stimulate its potential on complex tasks, the details of performing prompt engineering, establishing domain knowledge library, and implementing complex tasks are illustrated in this study. Moreover, the proposed framework is verified on two typical tasks: network alarm analysis and network performance optimization. The good response accuracies and semantic similarities of 2400 test situations exhibit the great potential of LLM in optical networks.

Context-Based e2e Autonomous Operation in B5G Networks.

The research and innovation related to fifth-generation (5G) networks that has been carried out in recent years has decided on the fundamentals of the smart slice in radio access networks (RANs), as well as the autonomous fixed network operation. One of the most challenging objectives of beyond 5G (B5G) and sixth-generation (6G) networks is the deployment of mechanisms that enable smart end-to-end (e2e) network operation, which is required for the achievement of the stringent service requirements of the envisioned use cases to be supported in the short term. Therefore, smart actions, such as dynamic capacity allocation, flexible functional split, and dynamic slice management need to be performed in tight coordination with the autonomous capacity management of the fixed transport network infrastructure. Otherwise, the benefits of smart slice operation (i.e., cost and energy savings while ensuring per-slice service requirements) might be cancelled due to uncoordinated autonomous fixed network operation. Notably, the transport network in charge of supporting slices from the user equipment (UE) to the core expands across access and metro fixed networks. The required coordination needs to be performed while keeping the privacy of the radio and fixed network domains, which is important in multi-tenant scenarios where both network segments are managed by different operators. In this paper, we propose a novel approach that explores the concept of context-aware network operation, where the slice control anticipates the aggregated and anonymized information of the expected slice operation that is sent to the fixed network orchestrator in an asynchronous way. The context is then used as the input for the artificial intelligence (AI)-based models used by the fixed network control for the predictive capacity management of optical connections in support of RAN slices. This context-aware network operation aims at enabling accurate and reliable autonomous fixed network operation under extremely dynamic traffic originated by smart RAN operation. The exhaustive numerical results show that slice context availability improves the benchmarking fixed network predictive methods (90% reduction in prediction maximum error) remarkably in the foreseen B5G scenarios, for both access and metro segments and in heterogeneous service demand scenarios. Moreover, context-aware network operation enables robust and efficient operation of optical networks in support of dense RAN cells (>32 base stations per cell), while the benchmarking methods fail to guarantee different operational objectives.

Pre- and post-processing techniques for reinforcement-learning-based routing and spectrum assignment in elastic optical networks

Research on the routing and spectrum assignment (RSA) problem has long been conducted with the aim of efficiently utilizing the frequency resources of optical networks. Given the recent progress in machine learning (ML) technology, it has been reported that the application of ML in various areas of optical network design, operation, and management has the potential to bring about new innovations such as autonomous optical network operation and highly accurate estimation of network conditions. With regard to the RSA problem, it is expected that an algorithm that achieves better accommodation efficiency than conventional heuristic methods can be realized by applying reinforcement learning (RL), which well supports training in a simulation environment. In this paper, we introduce and evaluate three techniques devised to apply RL more effectively to elastic optical networks (EONs): two pre-processing techniques called link-axis positional encoding (LPE) and slot-axis positional encoding (SPE) and a post-processing technique named the assignable boundary slot mask. First, we build a simple model in which the state data of the optical network, including frequency slot utilization, are input to the neural network of the RSA agent in RL and show that this model has difficulty outperforming the conventional heuristic RSA algorithm. Next, we build an RSA agent model with the proposed techniques and simulate the accommodation of dynamic optical paths to quantitatively demonstrate that the blocking probability can be reduced by 17% compared to the conventional heuristic.

Optical Networks Automation Overview: A Survey

The increasing demand for data has driven the advancement of optical networks from traditional architectures to more flexible, dynamic and efficient solutions. This includes technologies like flexgrid reconfigurable optical add-drop multiplexers (ROADMs), variable bandwidth transponders (VBTs) providing different modulation, coding schemes and baud rates. These advancements have brought about new challenges that concerns to the routing and spectrum allocation (RSA), fragmented spectrum, need for rapid and efficient channel restoration, and operation and maintenance management of optical networks. To address these challenges, a dynamic and flexible network requires a highly advanced network operational system (OS) capable of efficiently managing and allocating network resources. It relies on network abstraction, sensors, actuators, and software-defined networking (SDN) to enable algorithms, management, control, and decision-making. Improving the sensing capabilities of the network is crucial. Modern hardware and sensor technology can help forecast fiber breaks, equipment failures, and other potential issues in advance, allowing for proactive actions to be taken. Machine learning (ML) methods have been proposed in the literature to enhance the accuracy of quality of transmission (QoT) estimation, mitigate nonlinearities and provide decisions. This reduces the need for conservative design margins, maximizes the capacity of optical network systems and reduces the investment in infrastructure. Failure management is a critical aspect of optical networks. Providing early-warning and proactive protection is essential. This includes detecting failures, localizing them, identifying the root causes, and estimating their magnitude. Quick response to failures is vital to maintaining network reliability

Covert fault detection with imbalanced data using an improved autoencoder for optical networks

Covert faults are characterized by the performance parameters falling within the normal range, without any observable abnormalities. These types of faults pose a significant risk as they present no apparent warning signs of potential danger. Therefore, it is crucial to establish an efficient covert fault detection method to ensure the reliable and stable operation of optical networks. Data-driven technology, which reveals the internal relations and data patterns between the historical data by mining and analyzing the historical data, offers a new perspective for covert fault detection. However, equipment failures are extremely rare in real optical network systems, and the data imbalance of covert fault samples poses a challenge for standard machine learning classifiers in learning precise decision boundaries. To address this challenge, we propose a fault detection scheme based on an improved autoencoder for covert fault detection under data imbalance. The designed covert fault detection model exclusively utilizes normal samples during training and remains unaffected by data imbalance. Specifically, the model is specifically designed according to a number of encoder and decoder components to learn the normal sample data patterns in the latent space and detect covert faults based on the reconstruction errors in that space. To validate the proposed scheme, we conducted experiments using actual backbone data. According to the results, the detection accuracy and F1 score of the designed model on the test set were 0.9811 and 0.9527, and the false negative and false positive rates were 0.0026 and 0.0227, respectively. Furthermore, the visualization of the latent space reconstruction error principle for detecting covert faults was implemented using the principal component analysis dimension reduction and scatter plots.

Deep reinforcement learning for proactive spectrum defragmentation in elastic optical networks

The immense growth of Internet traffic calls for advanced techniques to enable the dynamic operation of optical networks, efficient use of spectral resources, and automation. In this paper, we investigate the proactive spectrum defragmentation (SD) problem in elastic optical networks and propose a novel deep reinforcement learning-based framework DeepDefrag to increase spectral usage efficiency. Unlike the conventional, often threshold-based heuristic algorithms that address a subset of the defragmentation-related tasks and have limited automation capabilities, DeepDefrag jointly addresses the three main aspects of the SD process: determining when to perform defragmentation, which connections to reconfigure, and which part of the spectrum to reallocate them to. By considering service attributes, the spectrum occupancy state expressed by several different fragmentation metrics, and the reconfiguration cost, DeepDefrag is able to consistently select appropriate reconfiguration actions over the network lifetime and adapt to changing conditions. Extensive simulation results reveal superior performance of the proposed scheme over a scenario with exhaustive defragmentation and a well-known benchmark heuristic from the literature, achieving lower blocking probability at a smaller defragmentation overhead.

RAMP: A flat nanosecond optical network and MPI operations for distributed deep learning systems

Distributed deep learning (DDL) systems strongly depend on network performance. Current electronic packet switched (EPS) network architectures and technologies suffer from variable diameter topologies, low-bisection bandwidth and over-subscription affecting completion time of communication and collective operations. We introduce a near-exascale, full-bisection bandwidth, all-to-all, single-hop, all-optical network architecture with nanosecond reconfiguration called RAMP, which supports large-scale distributed and parallel computing systems (12.8 Tbps per node for up to 65,536 nodes). For the first time, a custom RAMP-x MPI strategy and a network transcoder is proposed to run MPI collective operations across the optical circuit switched (OCS) network in a schedule-less and contention-less manner. RAMP achieves 7.6-171× speed-up in completion time across all MPI operations compared to realistic EPS and OCS counterparts. It can also deliver a 1.3-16× and 7.8-58× reduction in Megatron and DLRM training time respectively while offering 38-47× and 6.4-26.5× improvement in energy consumption and cost respectively.

Improved QoT estimations through refined signal power measurements and data-driven parameter optimizations in a disaggregated and partially loaded live production network

Accurate quality of transmission (QoT) estimations are essential enablers for future low-margin dynamic optical network operations. However, physical parameter measurement uncertainties and other intractable signal propagation effects degrade the accuracy of QoT estimation, especially in live production networks. The recent trend of network disaggregation further exacerbates the issue, and a vendor-agnostic accurate QoT estimator is much needed. In this paper, we study Gaussian-noise-model-based QoT estimation in a large-scale disaggregated and partially loaded live production network with monitored physical layer data spanning across 8 months. We first propose refining the signal power measurements by combining the inline amplifier and optical channel monitor (OCM) power measurements, followed by estimating the gain and noise power profiles of each inline amplifier, which in turn improves QoT estimation accuracy. We further introduce an optical multiplex section and frequency bias to the analytical model to incorporate intractable location-specific and spectral effects in the network and proposed data-driven parameter optimizations to learn the biases as well as erbium-doped fiber amplifier noise figures. The (mean, standard deviation) of the QoT estimation errors were reduced from (−0.1043, 0.6037) dB using average amplifier power and (−0.7875, 0.6337) dB using OCM power to (−0.0964, 0.4649) dB after input parameter refinements and were further reduced to (0.0046, 0.2377) dB with data-driven parameter optimization. The proposed methodologies are simple procedures that network operators can adopt to optimize analytical-model-based QoT estimators and/or serve as feature engineering procedures preceding machine-learning-based QoT in realistic disaggregated live production networks.

Channel power equalization based on joint optimization of EDFA and ROADM configuration in open optical network.

Deterioration of the signal-to-noise ratio (SNR) is an important challenge in ultra-long multi optical line system (OLS) optical transmission systems. The non-uniform gain and cascading of the Erbium-doped fiber amplifier (EDFA) lead to SNR deterioration in transmission systems. In this paper, we propose two channel power equalization methods based on joint optimization of EDFA and Reconfigurable optical add-drop multiplexer (ROADM) configurations: 1) reinforcement learning (RL)-based channel power equalization (RL-PE) and 2) covariance matrix adaptive evolution strategy (CMA-ES) channel power equalization (CMA-PE). The simulation results indicate that the power equalization effect was improved by 1.9 dB through the CMA-PE method, while the RL-PE method led to a 1.5 dB improvement in an ultra-long 80-channel 7-OLS transmission system.

Measurement-device-independent continuous variable quantum key distribution protocol operation in optical transport networks

Measurement-device-independent continuous variable quantum key distribution protocol operation in optical transport networks

Application of no-light fault prediction of PON based on deep learning method

The increased demand for broadband communication services has led to a wider deployment of fiber in the last mile. However, this expansion comes with the challenge of high maintenance costs for optical networks. To address this issue, there is a need for cost-efficient and reliable monitoring solutions that can detect and locate faults in different fiber access architectures. One specific concern in the operation of Passive Optical Networks (PONs) is the occurrence of no-light faults in the PON ports. To tackle this practical problem, we propose an intelligent scheme based on Long Short-Term Memory Networks (LSTMs) and Gated Recurrent Units (GRUs). These deep learning models are trained on historical data of the Optical Line Terminal (OLT) and PON behavior to predict future fault occurrences in real-time. To evaluate the performance of our proposed scheme, we conducted extensive experiments using real-world data. The results demonstrate the superiority of our approach over traditional methods and state-of-the-art models, as evidenced by improved accuracy, precision, recall, and F1 measure.

Machine learning enhancement of a digital twin for wavelength division multiplexing network performance prediction leveraging quality of transmission parameter refinement

Digital twins capable of quality of transmission (QoT) estimation and prediction are powerful tools for efficient design and operation of optical networks. In this paper, we employ machine learning techniques to enhance both the QoT estimation and prediction capabilities of an optical network digital twin. By leveraging a method to infer or refine the unknown lumped loss distributions and amplifier gain spectra for accurate characterization of the current optical network state, the accuracy of estimation and prediction is substantially improved. For QoT prediction, we further develop a novel neural-network architecture for erbium-doped fiber amplifiers that, after factory training on a single device and with fully loaded configurations only, generalizes to partially loaded configurations seen after deployment and to different gain and tilt settings and other physical devices of the same type. We combine refined parameters and a novel method that predicts the difference in per-channel power when individual services are added or removed or multiple services are lost due to a fiber cut. The impact of error amplification due to cascading of individual components’ models is shown to be reduced, yielding predictions that are substantially more accurate than simply cascaded predictions. As the network ages, the digital twin can be updated by retraining while using only information available in the current network state.

Link load prediction in an optical network with restoration mechanisms

Knowledge about the future link load is an essential issue for optical network operators, especially in the case of network failure and the restoration of affected traffic. In this paper, we study the dynamic operation of an optical network in its normal non-failure state and the situation of a link failure followed by the restoration process. Data obtained during the simulation are used for link load prediction. We propose dynamic methods for multistep link load prediction. The analyzed link load is expressed in two ways: the bitrate of the overall traffic allocated to the link and the number of frequency slots occupied. The proposed prediction methods are expanded to include additional mechanisms that improve the forecasting quality expressed by the mean absolute percentage error metric. We evaluate developed methods on a dataset collected using a representative European network topology with realistic traffic containing diverse types of network transmissions. In broad numerical experiments, we prove the high prediction quality of regression algorithms aided by the proposed additional features.

Enhancing meta slot space space wavelength to solve blocking case in elastic optical networks

Elastic optical networks (EONs) are a promising technology for the development of flexible and wideband communication systems. The (FSU) frequency slot unit concept is used in this technology to define the bandwidth unit in the frequency domain. The use of a specified number of consecutive FSUs allows bandwidth to be assigned to a data stream in a flexible manner. Nonblocking optical switching networks are required for a successful elastic optical network operation. This article concentrates on the Wavelength Space Wavelength,Space, Wavelength switching network topologies that were previously presented. In comparison to the other architectures, the SSW switching fabric requires fewer Spectrum converters (SCs). n relation to Wavelength Space Wavelength, a prior study clarified that using the meta-slot approach reduces hardware complexity, which is measured by the number of frequency slot unit. A frequency range with one or more (FSUs) is called a meta-slot. The previous study did not discuss how to optimize the meta-slot class sizes, despite the fact that its effectiveness depends on the sizes of the meta-slot classes. Additionally, the use of meta-slots for Space Space Wavelength was not taken into account in the earlier analysis. The optimization of meta-slot class sizes is looked into in this study, and it is shown that this optimization may be modeled as the shortest path issue. For Wavelength Space Wavelength, the previously reported nonblocking conditions are contrasted with the meta-slot scheme optimized using the shortest path model. The outcome attests to the optimized meta-slot scheme's superiority. The distribution of meta-slots among S-switches is also crucial for Space Space Wavelength. The assignment of meta-slots is modeled as a bin-packing problem in the study. Thus, a well-known bin-packing heuristic can be used to find the assignment that is close to ideal. The number of S-switches is calculated for using the optimized meta-slot scheme and previously known nonblocking conditions. The outcome confirms that the meta-slot scheme benefits both Space, Wavelength and Wavelength Space Wavelength.

Multiple attention mechanisms-driven component fault location in optical networks with network-wide monitoring data

Fault location is an essential part of optical network operation and maintenance, and network operators have expectations to achieve highly accurate and precise fault location for reducing the cost of fault recovery. However, due to the scale of such networks, the volume of monitoring data (MD) is quite large, which poses a great challenge for fault location. An attention mechanism is an effective way to focus on the important information from massive input for the current task, which originates from the study of human vision. Targeting component fault location in optical networks, we propose an attention mechanism-based strategy, which consists of a sequence attention mechanism (SAT), a channel attention mechanism (CAT), a graph attention mechanism (GAT), and a fully connected neural network (FCNN). SAT, CAT, and GAT are applied for link, node, and network representation, respectively, taking corresponding MD as input. The FCNN is responsible for analyzing the correlation between MD and completing the fault location decision. All three attention mechanisms can filter out the more critical MD, assisting the FCNN to make more accurate decisions. We compare the performance of the proposed strategy and artificial neural networks (ANNs) in partial telemetry scenarios. Simulation results indicate that our strategy outperforms ANNs with respect to the accuracy of fault location by focusing on more critical MD and achieves a maximum improvement by 5.6%. Moreover, its feasibility with real data is verified on an experimental testbed consisting of hybrid optical-electrical switching nodes. Extensive results show that our strategy has the potential to achieve highly accurate fault location in real networks.

Self-Attention Mechanism-Based Multi-Channel QoT Estimation in Optical Networks

It is essential to estimate the quality of transmission (QoT) of lightpaths before their establishment for efficient planning and operation of optical networks. Due to the nonlinear effect of fibers, the deployed lightpaths influence the QoT of each other; thus, multi-channel QoT estimation is necessary, which provides complete QoT information for network optimization. Moreover, the different interfering channels have different effects on the channel under test. However, the existing artificial-neural-network-based multi-channel QoT estimators (ANN-QoT-E) neglect the different effects of the interfering channels in their input layer, which affects their estimation accuracy severely. In this paper, we propose a self-attention mechanism-based multi-channel QoT estimator (SA-QoT-E) to improve the estimation accuracy of the ANN-QoT-E. In the SA-QoT-E, the input features are designed as a sequence of feature vectors of channels that route the same path, and the self-attention mechanism dynamically assigns weights to the feature vectors of interfering channels according to their effects on the channel under test. Moreover, a hyperparameter search method is used to optimize the SA-QoT-E. The simulation results show that, compared with the ANN-QoT-E, our proposed SA-QoT-E achieves higher estimation accuracy, and can be directly applied to the network wavelength expansion scenarios without retraining.

Nonlinear SNR estimation based on the data augmentation-assisted DNN with a small-scale dataset.

Fiber nonlinearity is one of the major impairments for long-haul transmission systems. Therefore, estimating the nonlinear signal-to-noise ratio (SNRNL) is indispensable to guarantee the quality of transmission and manage the operation of optical networks. The deep neural network (DNN) has been successfully applied for the SNRNL estimation. However, the performance substantially degrades, when the mega dataset is inaccessible. Here, we demonstrate an accurate SNRNL estimation based on the data augmentation (DA)-assisted DNN with a small-scale dataset in the frequency domain. When the 95-GBaud dual-polarization 16 quadrature amplitude modulation (DP-16QAM) signal with variable optical launch powers from -2 to 4-dBm is numerically transmitted from 80-km to 1520-km standard single-mode fiber (SSMF), we identify that, in comparison with traditional DNN scheme, the DA allows the reduction of the training dataset size by about 60% while keeping the same mean absolute error (MAE) as 0.2-dB of SNRNL estimation. Meanwhile, the DA-assisted DNN scheme can reduce the MAE by about 0.14-dB compared with the traditional DNN scheme, when both SNRNL estimation schemes use 100 raw datasets which contain 700 symbols. Due to these observations, the DA-assisted DNN scheme is promising for field-trial nonlinear SNR estimation, especially when the collection of mega datasets is inconvenient.

Photonic neuromorphic technologies in optical communications

Abstract Machine learning (ML) and neuromorphic computing have been enforcing problem-solving in many applications. Such approaches found fertile ground in optical communications, a technological field that is very demanding in terms of computational speed and complexity. The latest breakthroughs are strongly supported by advanced signal processing, implemented in the digital domain. Algorithms of different levels of complexity aim at improving data recovery, expanding the reach of transmission, validating the integrity of the optical network operation, and monitoring data transfer faults. Lately, the concept of reservoir computing (RC) inspired hardware implementations in photonics that may offer revolutionary solutions in this field. In a brief introduction, I discuss some of the established digital signal processing (DSP) techniques and some new approaches based on ML and neural network (NN) architectures. In the main part, I review the latest neuromorphic computing proposals that specifically apply to photonic hardware and give new perspectives on addressing signal processing in optical communications. I discuss the fundamental topologies in photonic feed-forward and recurrent network implementations. Finally, I review the photonic topologies that were initially tested for channel equalization benchmark tasks, and then in fiber transmission systems, for optical header recognition, data recovery, and modulation format identification.

Sdudy on Fault Analysis Method of Optical Transmission Network Based on Deep Learning

In order to ensure the efficient and stable operation of optical transmission network, timely and efficient fault diagnosis is very necessary. This paper studies the fault analysis method of optical transmission network based on deep learning, and summarizes the fault analysis method and research status of optical transmission network based on deep learning from two aspects of deep learning and fault diagnosis, for your reference.

Optical network design and analysis tools: A test of time

Telecom operators' infrastructure is sustained by optical communication networks that provide the means for exchanging large amounts of information, which is essential for many modern society needs. Optical networks are characterized by rapid breakthroughs in a variety of technologies. Relevantly, the last decade encompassed remarkable advances in optical networks’ subfields of signal processing, electronics, photonics, communications, protocols, and control-plane architectures. Hence, these advancements unlocked unprecedented transmission capacities, reconfigurability and programmability, entailing an evolution in the way which networks were designed, planned, and analyzed. In this paper, we review the historical status of optical planning and design tools by focusing on the major enabling technologies and relevant landmarks of the last decade(s). We begin by pinpointing the major breakthroughs in the optical data plane, estimation models capturing the transmission medium behavior and the control plane. We then distil the implications that these advancements entail in the landscape of optical network design and analysis tools, which commonly sit “on top” of the control plane or as a fully separated entity. Then, we speculate with our view for the future, in which automatic validation of optical network operations and dimensioning jointly with learning/artificial intelligence mechanisms will permit zero-touch optical networking: i.e. updating, provisioning, and upgrading network capacities, by means of automation with minimal human intervention. We conclude with a proposal of an architecture that encompasses data and control planes in a comprehensive manner for paving the way towards zero-touch optical networking.