Pub Date : 2026-01-24DOI: 10.1016/j.osn.2026.100852
S.A.H. Mohsan, Haoyu Huang, Yi Hao, Qian Li, H.Y. Fu
Recently, visible light communication (VLC) has garnered substantial interest as a promising complement to radio frequency (RF) communication due to low-cost, high data rate, huge bandwidth, RF-free interference, enhanced security, and license-free spectrum. However, owning to propagation characteristics of light, VLC is vulnerable to critical challenges in practical scenarios such as short bandwidth of light emitting diodes (LEDs), pointing errors, beam misalignment and inevitable occlusion. In this context, non-orthogonal multiple access (NOMA) and optical intelligent reflecting surface (OIRS) are two emerging techniques to overcome performance degradation of existing VLC systems. These burgeoning technologies offer unique potentials within the realm of VLC systems. The operating principle of NOMA is power allocation based on channel conditions and OIRS can smartly control the optical propagation channel, which motivate us to study OIRS-assisted NOMA-VLC systems for multi-user high-speed communications. Particularly, this study is primarily attributed to OIRS integration for exploiting reflected propagation, which significantly addresses the inherent signal blockage problems encountered in VLC systems. Furthermore, we have categorized the advantages of such unique integration as: the improvement of signal quality, the control of optical channel, the enhancement of communication security, and interference cancellation. In addition, we have discussed key factors pertaining to the tuning mechanism, channel characteristics, deployment, and performance metrics of OIRS is details. We have summarized key potentials and challenges for OIRS-assisted NOMA-VLC systems. Finally, we lay out a number of open research issues and trends for future breakthrough novelties in this domain.
{"title":"Optical Intelligent Reflecting Surface-assisted NOMA for VLC System: Fundamentals, Potentials, Open Research Issues, and Future Trends","authors":"S.A.H. Mohsan, Haoyu Huang, Yi Hao, Qian Li, H.Y. Fu","doi":"10.1016/j.osn.2026.100852","DOIUrl":"https://doi.org/10.1016/j.osn.2026.100852","url":null,"abstract":"Recently, visible light communication (VLC) has garnered substantial interest as a promising complement to radio frequency (RF) communication due to low-cost, high data rate, huge bandwidth, RF-free interference, enhanced security, and license-free spectrum. However, owning to propagation characteristics of light, VLC is vulnerable to critical challenges in practical scenarios such as short bandwidth of light emitting diodes (LEDs), pointing errors, beam misalignment and inevitable occlusion. In this context, non-orthogonal multiple access (NOMA) and optical intelligent reflecting surface (OIRS) are two emerging techniques to overcome performance degradation of existing VLC systems. These burgeoning technologies offer unique potentials within the realm of VLC systems. The operating principle of NOMA is power allocation based on channel conditions and OIRS can smartly control the optical propagation channel, which motivate us to study OIRS-assisted NOMA-VLC systems for multi-user high-speed communications. Particularly, this study is primarily attributed to OIRS integration for exploiting reflected propagation, which significantly addresses the inherent signal blockage problems encountered in VLC systems. Furthermore, we have categorized the advantages of such unique integration as: the improvement of signal quality, the control of optical channel, the enhancement of communication security, and interference cancellation. In addition, we have discussed key factors pertaining to the tuning mechanism, channel characteristics, deployment, and performance metrics of OIRS is details. We have summarized key potentials and challenges for OIRS-assisted NOMA-VLC systems. Finally, we lay out a number of open research issues and trends for future breakthrough novelties in this domain.","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"29 1","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-17DOI: 10.1016/j.osn.2026.100851
Wessam M. Salama , Moustafa H. Aly , Samah Alshathri , Walid El-Shafai
Underwater Visible Light Communication (UVLC) is a promising solution for short-range underwater applications due to its high data rates and spectral efficiency. This paper presents an intelligent bidirectional UVLC system that incorporates real-time beam adaptation and intelligent modulation control to enhance communication robustness. The system integrates Software-Defined Radios (SDRs), an ultra-wide field-of-view photodetector, and a 470 nm Li-Fi transmitter, all coordinated through a simulated servo-driven beam-steering mechanism. This configuration enables dynamic adjustment of beam divergence angles to maximize received optical power and maintain stable communication across diverse underwater environments. Pulse Position Modulation (PPM) is employed, and a comparative evaluation between 4-PPM and 8-PPM is conducted. While 8-PPM provides higher data throughput, 4-PPM demonstrates superior noise robustness and lower Bit Error Rate (BER), particularly under misalignment and degraded channel conditions. A key contribution of this work is DenseFormer, a hybrid Deep Learning (DL) architecture that combines the global contextual modeling capability of Transformer encoders with the spatial feature extraction strengths of DenseNet169. Trained on a large synthetic dataset, DenseFormer effectively predicts the complex interdependencies among angular position, divergence angle, modulation type, and received optical power, thereby enabling data-driven BER and power optimization. Comparative analysis shows that DenseFormer substantially outperforms the baseline DenseNet169 model in prediction accuracy and generalization. Simulation results demonstrate up to 98 % BER reduction in Non-Line-of-Sight (NLoS) scenarios, 50 % BER reduction at critical misalignment angles, and a 47.4 % increase in received power in low-signal angular regions. Overall, this work demonstrates that combining deep learning, SDR-based Li-Fi transceivers, and adaptive beam control enables durable, high-throughput, and resilient UVLC suitable for next-generation underwater communication systems.
{"title":"DenseFormer-UVLC: Intelligent hybrid deep learning framework for adaptive beam steering and modulation in underwater visible light communication","authors":"Wessam M. Salama , Moustafa H. Aly , Samah Alshathri , Walid El-Shafai","doi":"10.1016/j.osn.2026.100851","DOIUrl":"10.1016/j.osn.2026.100851","url":null,"abstract":"<div><div>Underwater Visible Light Communication (UVLC) is a promising solution for short-range underwater applications due to its high data rates and spectral efficiency. This paper presents an intelligent bidirectional UVLC system that incorporates real-time beam adaptation and intelligent modulation control to enhance communication robustness. The system integrates Software-Defined Radios (SDRs), an ultra-wide field-of-view photodetector, and a 470 nm Li-Fi transmitter, all coordinated through a simulated servo-driven beam-steering mechanism. This configuration enables dynamic adjustment of beam divergence angles to maximize received optical power and maintain stable communication across diverse underwater environments. Pulse Position Modulation (PPM) is employed, and a comparative evaluation between 4-PPM and 8-PPM is conducted. While 8-PPM provides higher data throughput, 4-PPM demonstrates superior noise robustness and lower Bit Error Rate (BER), particularly under misalignment and degraded channel conditions. A key contribution of this work is DenseFormer, a hybrid Deep Learning (DL) architecture that combines the global contextual modeling capability of Transformer encoders with the spatial feature extraction strengths of DenseNet169. Trained on a large synthetic dataset, DenseFormer effectively predicts the complex interdependencies among angular position, divergence angle, modulation type, and received optical power, thereby enabling data-driven BER and power optimization. Comparative analysis shows that DenseFormer substantially outperforms the baseline DenseNet169 model in prediction accuracy and generalization. Simulation results demonstrate up to 98 % BER reduction in Non-Line-of-Sight (NLoS) scenarios, 50 % BER reduction at critical misalignment angles, and a 47.4 % increase in received power in low-signal angular regions. Overall, this work demonstrates that combining deep learning, SDR-based Li-Fi transceivers, and adaptive beam control enables durable, high-throughput, and resilient UVLC suitable for next-generation underwater communication systems.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"60 ","pages":"Article 100851"},"PeriodicalIF":3.1,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-09DOI: 10.1016/j.osn.2025.100841
Huanlin Liu , Dongxin An , Yong Chen , Xiang Chen , Jiachen Zou , Haonan Chen , Bing Ma
Elastic optical networks (EONs) face with more intense resource competition and difficulties in ensuring service quality for time-varying services. This study uses deep Q-network to optimize routing, modulation and spectrum allocation (DQN-RMSA) for time-varying services. The DQN is introduced to combine the deep reinforcement learning (DRL) framework and Q-network to manage RMSA by sliding windows for time-varying services. Specifically, a priority-aware state representation is adopted to integrate time-varying service characteristics and network resource state capturing by the DRL agent. An enhanced prioritized experience replay mechanism with a Sumtree structure is used to accelerate convergence by prioritizing critical learning samples. Moreover, a delayed action is designed by using Markov decision process (MDP) to monitor state transition for improving the performance of RMSA, and the delay action is reflected in the reward function to balance delay and resource utilization in terms of decreasing blocking probability. Simulation results demonstrate that DQN-RMSA can achieve superior performance over traditional heuristic and other DRL methods, with notable improvements in service blocking probability, delay tolerance, and spectrum utilization. The DQN-RMSA can adapt to fluctuations in time-varying service, highlighting its potential for achieving self-optimization and robust resource allocation in dynamic environments.
{"title":"Intelligent service RMSA in elastic optical networks: deep Q-networks","authors":"Huanlin Liu , Dongxin An , Yong Chen , Xiang Chen , Jiachen Zou , Haonan Chen , Bing Ma","doi":"10.1016/j.osn.2025.100841","DOIUrl":"10.1016/j.osn.2025.100841","url":null,"abstract":"<div><div>Elastic optical networks (EONs) face with more intense resource competition and difficulties in ensuring service quality for time-varying services. This study uses deep Q-network to optimize routing, modulation and spectrum allocation (DQN-RMSA) for time-varying services. The DQN is introduced to combine the deep reinforcement learning (DRL) framework and Q-network to manage RMSA by sliding windows for time-varying services. Specifically, a priority-aware state representation is adopted to integrate time-varying service characteristics and network resource state capturing by the DRL agent. An enhanced prioritized experience replay mechanism with a Sumtree structure is used to accelerate convergence by prioritizing critical learning samples. Moreover, a delayed action is designed by using Markov decision process (MDP) to monitor state transition for improving the performance of RMSA, and the delay action is reflected in the reward function to balance delay and resource utilization in terms of decreasing blocking probability. Simulation results demonstrate that DQN-RMSA can achieve superior performance over traditional heuristic and other DRL methods, with notable improvements in service blocking probability, delay tolerance, and spectrum utilization. The DQN-RMSA can adapt to fluctuations in time-varying service, highlighting its potential for achieving self-optimization and robust resource allocation in dynamic environments.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"59 ","pages":"Article 100841"},"PeriodicalIF":3.1,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145731343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-29DOI: 10.1016/j.osn.2025.100831
Andrei N. Ribeiro , Fabrício R. Lobato , Moisés F. Silva , João C.W.A. Costa
The continuous emergence of data-intensive online applications has required efforts to ensure the reliability of high data transmissions provided by optical networks. Considering that these systems are susceptible to faults that can affect their functioning, providing practical and accurate fault management approaches becomes vital. Therefore, this paper presents a semi-supervised approach based on an autoencoder (AE) model for detecting and localizing faults in optical networks. Since data from fault conditions is often absent in practical scenarios, a semi-supervised training strategy is applied, using only data from normal operating conditions. We can track fault occurrences based on the reconstruction error (RE) generated by the AE, which can serve as fault indicators (FIs), as data from normal and faulty conditions will produce different REs. Moreover, the proposed approach can localize faults by detecting them individually for each piece of equipment in the network. A dataset derived from an optical testbed is used for performance evaluation, focusing on accuracy and Type I and Type II errors. Results demonstrate that the standard AE outperformed PCA and other semi-supervised methods, including Variational AE and LSTM-AE, achieving an average fault detection accuracy of approximately 92.05%, with Type I and Type II error rates of about 7% and 0.9%, respectively.
{"title":"A semi-supervised ML-driven strategy for fault detection and localization in optical networks","authors":"Andrei N. Ribeiro , Fabrício R. Lobato , Moisés F. Silva , João C.W.A. Costa","doi":"10.1016/j.osn.2025.100831","DOIUrl":"10.1016/j.osn.2025.100831","url":null,"abstract":"<div><div>The continuous emergence of data-intensive online applications has required efforts to ensure the reliability of high data transmissions provided by optical networks. Considering that these systems are susceptible to faults that can affect their functioning, providing practical and accurate fault management approaches becomes vital. Therefore, this paper presents a semi-supervised approach based on an autoencoder (AE) model for detecting and localizing faults in optical networks. Since data from fault conditions is often absent in practical scenarios, a semi-supervised training strategy is applied, using only data from normal operating conditions. We can track fault occurrences based on the reconstruction error (RE) generated by the AE, which can serve as fault indicators (FIs), as data from normal and faulty conditions will produce different REs. Moreover, the proposed approach can localize faults by detecting them individually for each piece of equipment in the network. A dataset derived from an optical testbed is used for performance evaluation, focusing on accuracy and Type I and Type II errors. Results demonstrate that the standard AE outperformed PCA and other semi-supervised methods, including Variational AE and LSTM-AE, achieving an average fault detection accuracy of approximately 92.05%, with Type I and Type II error rates of about 7% and 0.9%, respectively.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"59 ","pages":"Article 100831"},"PeriodicalIF":3.1,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145613794","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-04DOI: 10.1016/j.osn.2025.100830
Ting-Yi Cheng, Der-Rong Din
We study Semi-Filterless Optical Networks (Semi-FON), where a limited number of Wavelength Selective Switch (WSS) nodes are deployed incrementally to improve spectrum utilization at low cost. We address two coupled problems under new WSS placements: (i) reconstructing the fiber-tree substrate and (ii) allocating spectrum to minimize the maximum spectrum-slot usage. We propose two complementary reconstruction strategies — Subtree and Single-Link — combined with a greedy spectrum-minimization allocator. The methods support single- and two-level FON architectures and guarantee connectivity for arbitrary source–destination pairs.
Simulations on connected fixed topologies and several random topologies across multiple sizes and edge densities (under fixed and random traffic) reveal a robust two-phase behavior: early selections deliver sharp objective reductions, followed by diminishing returns. Subtree consistently achieves lower objectives with only a few WSS placements and reaches a plateau quickly, making it attractive under tight budgets. With larger budgets, Single-Link can continue making small but steady improvements and may eventually match or surpass Subtree. Runtime scales more strongly with network size than density, but remains practical for the evaluated regimes. Overall, Semi-FON with the proposed strategies substantially reduces peak spectrum usage while enabling cost-aware, incremental deployment.
{"title":"Incremental placement of wavelength-selective switching nodes in Semi-Filterless Optical Networks","authors":"Ting-Yi Cheng, Der-Rong Din","doi":"10.1016/j.osn.2025.100830","DOIUrl":"10.1016/j.osn.2025.100830","url":null,"abstract":"<div><div>We study Semi-Filterless Optical Networks (Semi-FON), where a limited number of Wavelength Selective Switch (WSS) nodes are deployed incrementally to improve spectrum utilization at low cost. We address two coupled problems under new WSS placements: (i) reconstructing the fiber-tree substrate and (ii) allocating spectrum to minimize the maximum spectrum-slot usage. We propose two complementary reconstruction strategies — Subtree and Single-Link — combined with a greedy spectrum-minimization allocator. The methods support single- and two-level FON architectures and guarantee connectivity for arbitrary source–destination pairs.</div><div>Simulations on connected fixed topologies and several random topologies across multiple sizes and edge densities (under fixed and random traffic) reveal a robust two-phase behavior: early selections deliver sharp objective reductions, followed by diminishing returns. Subtree consistently achieves lower objectives with only a few WSS placements and reaches a plateau quickly, making it attractive under tight budgets. With larger budgets, Single-Link can continue making small but steady improvements and may eventually match or surpass Subtree. Runtime scales more strongly with network size than density, but remains practical for the evaluated regimes. Overall, Semi-FON with the proposed strategies substantially reduces peak spectrum usage while enabling cost-aware, incremental deployment.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"58 ","pages":"Article 100830"},"PeriodicalIF":3.1,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145441600","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-13DOI: 10.1016/j.osn.2025.100829
Buniechukwu Njoku , Milad Ghadimi , Swaraj Shekhar Nande , Akhmadjon Rajabov , Ming Yin , Ju Hoon Kim , Ernest Scholtz , Muhammad Idham Habibie , Bassem Arar , Caspar Hopfmann , Riccardo Bassoli , Frank H.P. Fitzek
This work presents a hybrid quantum–classical architecture that converts entanglement-derived resources including ultra-precise timing, private/common randomness and certified entropy into first-class primitives for 5G networks. Semiconductor quantum-dot sources and correlation electronics distribute photon-pair coincidences, while a layered protocol stack exposes the resulting services to 5G core functions. As a proof-of-concept, we embed a quantum-random-number generator (QRNG) into the 5G Authentication and Key Agreement (AKA) procedure, replacing the pseudo-random RAND generator in Open5GS with an HTTP-served QRNG feed. A simulation of sequential UE attachments shows that the QRNG path increases cumulative control-plane latency by and produces isolated Round-Trip-Time (RTT) spikes, yet leaves data-plane efficiency largely unchanged: delivery rate matches the baseline while a flatter pacing profile lowers the risk of queue build-up. These results quantify the performance trade-off of QRNG-enhanced 5G and motivate tighter in-process or hardware QRNG integration in future quantum-enabled mobile networks.
{"title":"Quantum entanglement resource utilization in quantum-classical networking","authors":"Buniechukwu Njoku , Milad Ghadimi , Swaraj Shekhar Nande , Akhmadjon Rajabov , Ming Yin , Ju Hoon Kim , Ernest Scholtz , Muhammad Idham Habibie , Bassem Arar , Caspar Hopfmann , Riccardo Bassoli , Frank H.P. Fitzek","doi":"10.1016/j.osn.2025.100829","DOIUrl":"10.1016/j.osn.2025.100829","url":null,"abstract":"<div><div>This work presents a hybrid quantum–classical architecture that converts entanglement-derived resources including ultra-precise timing, private/common randomness and certified entropy into first-class primitives for 5G networks. Semiconductor quantum-dot sources and correlation electronics distribute photon-pair coincidences, while a layered protocol stack exposes the resulting services to 5G core functions. As a proof-of-concept, we embed a quantum-random-number generator (QRNG) into the 5G Authentication and Key Agreement (AKA) procedure, replacing the pseudo-random <span>RAND</span> generator in Open5GS with an HTTP-served QRNG feed. A simulation of sequential UE attachments shows that the QRNG path increases cumulative control-plane latency by <span><math><mrow><mo>≈</mo><mspace></mspace><mn>12</mn><mtext>%</mtext></mrow></math></span> and produces isolated Round-Trip-Time (RTT) spikes, yet leaves data-plane efficiency largely unchanged: delivery rate matches the baseline while a flatter pacing profile lowers the risk of queue build-up. These results quantify the performance trade-off of QRNG-enhanced 5G and motivate tighter in-process or hardware QRNG integration in future quantum-enabled mobile networks.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"58 ","pages":"Article 100829"},"PeriodicalIF":3.1,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145314944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-12DOI: 10.1016/j.osn.2025.100828
Yunxin Lv, Pengcheng Li, Meihua Bi, Yanrong Zhai, Hao Chi
A low-latency and energy-efficient resource allocation scheme is proposed and investigated for time division multiplexing passive optical networks (TDM-PONs) having heterogeneous optical network unit (ONU) propagation delays. By (a) properly postponing the uplink transmissions of ONUs and (b) in each polling cycle, searching the postponed uplink transmissions and selecting the next ONU to poll based on the time overlap between uplink and downlink transmissions as well as the resultant channel idle time, the proposed scheme minimizes both the overall operating time of each ONU and the channel idle time caused by heterogeneous ONU propagation delays. Consequently, available sleep time for each ONU is ensured, and high channel utilization efficiency is secured, thereby resulting in the energy-efficient and low-latency performance of the proposed scheme. Simulation results indicate that for a 10G-EPON with ONU propagation delays distributed in the range from 25 μs to 200 μs, the proposed scheme can significantly reduce the average packet delay compared to the conventional schemes, and is more energy-efficient in delivering data under a 1 ms delay constraint.
{"title":"Uplink postponing with downlink matching: A low-latency and energy-efficient resource allocation scheme for TDM-PONs having heterogeneous ONU propagation delays","authors":"Yunxin Lv, Pengcheng Li, Meihua Bi, Yanrong Zhai, Hao Chi","doi":"10.1016/j.osn.2025.100828","DOIUrl":"10.1016/j.osn.2025.100828","url":null,"abstract":"<div><div>A low-latency and energy-efficient resource allocation scheme is proposed and investigated for time division multiplexing passive optical networks (TDM-PONs) having heterogeneous optical network unit (ONU) propagation delays. By (a) properly postponing the uplink transmissions of ONUs and (b) in each polling cycle, searching the postponed uplink transmissions and selecting the next ONU to poll based on the time overlap between uplink and downlink transmissions as well as the resultant channel idle time, the proposed scheme minimizes both the overall operating time of each ONU and the channel idle time caused by heterogeneous ONU propagation delays. Consequently, available sleep time for each ONU is ensured, and high channel utilization efficiency is secured, thereby resulting in the energy-efficient and low-latency performance of the proposed scheme. Simulation results indicate that for a 10G-EPON with ONU propagation delays distributed in the range from 25 μs to 200 μs, the proposed scheme can significantly reduce the average packet delay compared to the conventional schemes, and is more energy-efficient in delivering data under a 1 ms delay constraint.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"58 ","pages":"Article 100828"},"PeriodicalIF":3.1,"publicationDate":"2025-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145314872","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The introduction of all-optical switching in datacenter internal networks (DCNs) marks a significant step toward overcoming the limitations of traditional electronic switching. However, challenges such as limited optical switch port counts and slow reconfiguration speeds demand new architectural approaches. In this paper, we propose a DCN fabric based on a “Lean” optical switch design that offers scalable, partially configurable switching. This architecture supports fast reconfiguration suitable for real-time scheduling and network control, while optimizing hardware complexity. To achieve this, we relax the non-blocking network constraint and adopt partially configurable switching modules, which reduce control complexity by limiting the scheduler's configuration space. We compare the proposed Lean network against two established optical DCN architectures: RotorNet, which relies on fully distributed control with minimal configurability, and optical Folded-Clos, which employs centralized control and fully reconfigurable optical crossbars. These architectures span different design trade-offs in terms of control and switching complexity. Our Lean design strikes a balance between the two, combining their respective benefits—offering moderate configurability, fast reconfiguration, and reduced control overhead. We analyze the crosspoint and scheduling complexities of all three architectures and evaluate their throughput and latency performance through packet level simulations.
{"title":"A Lean and fast optical datacenter internal network fabric with partial configurability","authors":"Konstantinos Kontodimas , Konstantinos Christodoulopoulos , Emmanouel Varvarigos","doi":"10.1016/j.osn.2025.100827","DOIUrl":"10.1016/j.osn.2025.100827","url":null,"abstract":"<div><div>The introduction of all-optical switching in datacenter internal networks (DCNs) marks a significant step toward overcoming the limitations of traditional electronic switching. However, challenges such as limited optical switch port counts and slow reconfiguration speeds demand new architectural approaches. In this paper, we propose a DCN fabric based on a “Lean” optical switch design that offers scalable, partially configurable switching. This architecture supports fast reconfiguration suitable for real-time scheduling and network control, while optimizing hardware complexity. To achieve this, we relax the non-blocking network constraint and adopt partially configurable switching modules, which reduce control complexity by limiting the scheduler's configuration space. We compare the proposed Lean network against two established optical DCN architectures: RotorNet, which relies on fully distributed control with minimal configurability, and optical Folded-Clos, which employs centralized control and fully reconfigurable optical crossbars. These architectures span different design trade-offs in terms of control and switching complexity. Our Lean design strikes a balance between the two, combining their respective benefits—offering moderate configurability, fast reconfiguration, and reduced control overhead. We analyze the crosspoint and scheduling complexities of all three architectures and evaluate their throughput and latency performance through packet level simulations.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"58 ","pages":"Article 100827"},"PeriodicalIF":3.1,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145314850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-30DOI: 10.1016/j.osn.2025.100826
Yuta Mori, Maiko Shigeno
Optical networks play an essential role as an infrastructure supporting advanced information and communication technology. Elastic Optical Networks (EONs) have been designed to improve frequency resource utilization efficiency and enable flexible and efficient network operations. In EONs, the Routing and Spectrum Allocation (RSA) problem — which involves determining routes of transmission and allocating frequency resources — is a critical issue for efficient network operation. Traditionally, the k-Shortest Paths (kSP) approach has been widely used for candidate transmission routes. However, kSP often leads to the concentration of transmission on specific links, creating bottlenecks for resource allocation. To address this challenge, we propose three novel candidate path selection methods for the RSA problem: (1) ensuring path diversity through similarity constraints via 0–1 integer programming, (2) clustering based on path similarity, and (3) suppressing link sharing through node pair grouping. Numerical experiments performed to assure the effectiveness of our proposed methods, and confirmed that introducing path diversity can yield high-quality solutions to the RSA problem under specific scenarios.
{"title":"Selection of candidate paths based on diversity in resource allocation problems on optical networks","authors":"Yuta Mori, Maiko Shigeno","doi":"10.1016/j.osn.2025.100826","DOIUrl":"10.1016/j.osn.2025.100826","url":null,"abstract":"<div><div>Optical networks play an essential role as an infrastructure supporting advanced information and communication technology. Elastic Optical Networks (EONs) have been designed to improve frequency resource utilization efficiency and enable flexible and efficient network operations. In EONs, the Routing and Spectrum Allocation (RSA) problem — which involves determining routes of transmission and allocating frequency resources — is a critical issue for efficient network operation. Traditionally, the k-Shortest Paths (kSP) approach has been widely used for candidate transmission routes. However, kSP often leads to the concentration of transmission on specific links, creating bottlenecks for resource allocation. To address this challenge, we propose three novel candidate path selection methods for the RSA problem: (1) ensuring path diversity through similarity constraints via 0–1 integer programming, (2) clustering based on path similarity, and (3) suppressing link sharing through node pair grouping. Numerical experiments performed to assure the effectiveness of our proposed methods, and confirmed that introducing path diversity can yield high-quality solutions to the RSA problem under specific scenarios.</div></div>","PeriodicalId":54674,"journal":{"name":"Optical Switching and Networking","volume":"58 ","pages":"Article 100826"},"PeriodicalIF":3.1,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145261960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：