This Special Issue on Intelligent Networks at the Edge contains a collection of invitation-only extensions based on papers presented at the Future Optical Networks and Communications (FONC) Symposium in the IEEE Future Network World Forum (FNWF), 13–15 November 2023, Baltimore, Maryland, USA. We present a brief introduction followed by an overview of the topics covered in the papers.
{"title":"Introduction to the JOCN Special Issue on Intelligent Networks at the Edge","authors":"Daniel Kilper;Lena Wosinska;Zuqing Zhu","doi":"10.1364/JOCN.541552","DOIUrl":"https://doi.org/10.1364/JOCN.541552","url":null,"abstract":"This Special Issue on Intelligent Networks at the Edge contains a collection of invitation-only extensions based on papers presented at the Future Optical Networks and Communications (FONC) Symposium in the IEEE Future Network World Forum (FNWF), 13–15 November 2023, Baltimore, Maryland, USA. We present a brief introduction followed by an overview of the topics covered in the papers.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"INE1-INE1"},"PeriodicalIF":4.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10680855","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142235830","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ensuring the secure and reliable operation of optical networks is crucial for various societal functions. However, optical network infrastructures are susceptible to unauthorized interception, posing a significant security risk at the physical layer. This necessitates the development of effective detection and localization methods of eavesdropping events. To address this challenge, we present a clustering-based method and a comprehensive eavesdropping diagnosis framework tailored for wavelength division multiplexing (WDM) systems. The framework is designed to handle diverse eavesdropping scenarios, including dynamic detection, classification, and localization of eavesdropping events. To mitigate the data dependency issue while detecting and localizing eavesdropping events, we propose a clustering algorithm utilizing basic optical performance monitoring (OPM) data, thus eliminating the need for sophisticated measurement equipment. A coarse localization requires only the OPM data from the receiver, while a finer localization requires the power monitoring data at all nodes as the input. The feasibility of the proposed scheme is validated using simulation-generated data, in which single and multiple eavesdropping can be detected and localized with a 100% label matching rate. Single-point eavesdropping detection and localization are experimentally validated with data collected from a fiber transmission system comprising three spans of 40 km each. Coarse localization with a 99.79% label matching rate and fine localization with 100% accuracy is achieved. As expected, experimental data shows a less concentrated distribution than the simulated data, which leads to inferior clustering results.
{"title":"Cluster-based unsupervised method for eavesdropping detection and localization in WDM systems","authors":"Haokun Song;Rui Lin;Lena Wosinska;Paolo Monti;Mingrui Zhang;Yuyuan Liang;Yajie Li;Jie Zhang","doi":"10.1364/JOCN.531696","DOIUrl":"10.1364/JOCN.531696","url":null,"abstract":"Ensuring the secure and reliable operation of optical networks is crucial for various societal functions. However, optical network infrastructures are susceptible to unauthorized interception, posing a significant security risk at the physical layer. This necessitates the development of effective detection and localization methods of eavesdropping events. To address this challenge, we present a clustering-based method and a comprehensive eavesdropping diagnosis framework tailored for wavelength division multiplexing (WDM) systems. The framework is designed to handle diverse eavesdropping scenarios, including dynamic detection, classification, and localization of eavesdropping events. To mitigate the data dependency issue while detecting and localizing eavesdropping events, we propose a clustering algorithm utilizing basic optical performance monitoring (OPM) data, thus eliminating the need for sophisticated measurement equipment. A coarse localization requires only the OPM data from the receiver, while a finer localization requires the power monitoring data at all nodes as the input. The feasibility of the proposed scheme is validated using simulation-generated data, in which single and multiple eavesdropping can be detected and localized with a 100% label matching rate. Single-point eavesdropping detection and localization are experimentally validated with data collected from a fiber transmission system comprising three spans of 40 km each. Coarse localization with a 99.79% label matching rate and fine localization with 100% accuracy is achieved. As expected, experimental data shows a less concentrated distribution than the simulated data, which leads to inferior clustering results.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"F52-F61"},"PeriodicalIF":4.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141929432","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Higher-speed passive optical networks (i.e., ITU-T 50G PONs) are envisioned to support various time-sensitive services with diverse quality of service (QoS) requirements (e.g., latency and jitter). To achieve this goal, higher-speed PON introduces multiple queues in each transmission container (T-CONT) so that each queue can carry a type of time-sensitive traffic. However, in the existing dynamic bandwidth allocation (DBA) schemes, the DBA engine in the optical line terminal (OLT) allocates bandwidth to four types of T-CONTs based on their buffer occupancies and priorities without considering queue status (e.g., the amount of data frames in queues) and services’ QoS requirements. To support multiple services belonging to the same T-CONT, the traffic scheduler in each optical network unit (ONU) further assigns the bandwidth of T-CONTs to their associated queues by using priority-based scheduling strategies. In this way, the bandwidth is allocated to queues by the DBA engine and traffic scheduler independently, which cannot guarantee network performance and meet the stringent QoS requirements of time-sensitive services. To solve this problem, we propose a DBA mechanism based on flexible queue management (FQM) to enable direct allocation of bandwidth to the queues under their QoS requirements. The proposed FQM mechanism enables the DBA engine to obtain the arrival time and QoS requirements of data frames in different queues as well as queue status based on the existing report/grant mechanism. By using these parameters, the required bandwidth for each queue in the next polling cycle can be calculated. Then, the DBA engine allocates bandwidth to these queues according to their bandwidth requests and priorities periodically. Simulation results show that the proposed scheme outperforms two benchmarks in the aspects of meeting time-sensitive services’ diverse QoS requirements, even when the traffic load is high.
{"title":"Flexible-queue-management-based bandwidth allocation in higher-speed PONs","authors":"Jun Li;Guanlun Sun;Xiang Lu;Rui Lin;Lena Wosinska","doi":"10.1364/JOCN.525843","DOIUrl":"10.1364/JOCN.525843","url":null,"abstract":"Higher-speed passive optical networks (i.e., ITU-T 50G PONs) are envisioned to support various time-sensitive services with diverse quality of service (QoS) requirements (e.g., latency and jitter). To achieve this goal, higher-speed PON introduces multiple queues in each transmission container (T-CONT) so that each queue can carry a type of time-sensitive traffic. However, in the existing dynamic bandwidth allocation (DBA) schemes, the DBA engine in the optical line terminal (OLT) allocates bandwidth to four types of T-CONTs based on their buffer occupancies and priorities without considering queue status (e.g., the amount of data frames in queues) and services’ QoS requirements. To support multiple services belonging to the same T-CONT, the traffic scheduler in each optical network unit (ONU) further assigns the bandwidth of T-CONTs to their associated queues by using priority-based scheduling strategies. In this way, the bandwidth is allocated to queues by the DBA engine and traffic scheduler independently, which cannot guarantee network performance and meet the stringent QoS requirements of time-sensitive services. To solve this problem, we propose a DBA mechanism based on flexible queue management (FQM) to enable direct allocation of bandwidth to the queues under their QoS requirements. The proposed FQM mechanism enables the DBA engine to obtain the arrival time and QoS requirements of data frames in different queues as well as queue status based on the existing report/grant mechanism. By using these parameters, the required bandwidth for each queue in the next polling cycle can be calculated. Then, the DBA engine allocates bandwidth to these queues according to their bandwidth requests and priorities periodically. Simulation results show that the proposed scheme outperforms two benchmarks in the aspects of meeting time-sensitive services’ diverse QoS requirements, even when the traffic load is high.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"F40-F51"},"PeriodicalIF":4.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141928478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Leitao Wang;Zhiyong Xu;Jingyuan Wang;Jiyong Zhao;Yimin Wang;Jianhua Li
Ultraviolet (UV) optical communication can eliminate aiming, tracking, and pointing systems in conventional free-space optical (FSO) communications, rendering it more suitable for small and micro platforms. Currently, efficient medium access control (MAC) layer protocols for UV optical communication are lacking. To further improve the channel utilization of UV networks, an ultraviolet lossless contention MAC protocol with dynamic bandwidth allocation (UVLLC-DBA) based on optical power superposition was proposed. Simultaneously, the protocol employed token buckets to allocate bandwidth for different services, thereby enabling differentiated services. The simulation results show that the proposed protocol can achieve dynamic bandwidth allocation and has nearly 100% channel utilization and acceptable delay, providing a reference for the development of a UV MAC protocol.
紫外（UV）光通信可以消除传统自由空间光通信（FSO）中的瞄准、跟踪和指向系统，因此更适用于小型和微型平台。目前，紫外光通信还缺乏高效的介质访问控制（MAC）层协议。为进一步提高紫外网络的信道利用率，提出了一种基于光功率叠加的动态带宽分配紫外无损争用 MAC 协议（UVLLC-DBA）。同时，该协议采用令牌桶为不同服务分配带宽，从而实现了差异化服务。仿真结果表明，所提出的协议可实现动态带宽分配，信道利用率接近 100%，延迟可接受，为开发 UV MAC 协议提供了参考。
{"title":"Lossless contention ultraviolet MAC protocol based on dynamic bandwidth allocation","authors":"Leitao Wang;Zhiyong Xu;Jingyuan Wang;Jiyong Zhao;Yimin Wang;Jianhua Li","doi":"10.1364/JOCN.523536","DOIUrl":"https://doi.org/10.1364/JOCN.523536","url":null,"abstract":"Ultraviolet (UV) optical communication can eliminate aiming, tracking, and pointing systems in conventional free-space optical (FSO) communications, rendering it more suitable for small and micro platforms. Currently, efficient medium access control (MAC) layer protocols for UV optical communication are lacking. To further improve the channel utilization of UV networks, an ultraviolet lossless contention MAC protocol with dynamic bandwidth allocation (UVLLC-DBA) based on optical power superposition was proposed. Simultaneously, the protocol employed token buckets to allocate bandwidth for different services, thereby enabling differentiated services. The simulation results show that the proposed protocol can achieve dynamic bandwidth allocation and has nearly 100% channel utilization and acceptable delay, providing a reference for the development of a UV MAC protocol.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"941-955"},"PeriodicalIF":4.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142173937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a routing, modulation, spectrum, and core assignment (RMSCA) algorithm for space-division-multiplexing-based elastic optical networks (SDM-EONs) comprising multi-core links. A network state-dependent route and core selection method is proposed using a deep neural network (DNN) classifier. The DNN is trained using a metaheuristic optimization algorithm to predict lightpath suitability, considering the quality of transmission and resource availability. Physical layer impairments, including inter-core crosstalk, amplified spontaneous emission, and Kerr fiber nonlinearities, are considered, and a random forest (RF)-based link noise estimator is proposed. A feature importance selection analysis is provided for all the features considered for the DNN classifier and the RF link noise estimator. The proposed machine-learning-enabled RMSCA approach is evaluated on three network topologies, USNET, NSFNET, and COST-239 with 7-core and 12-core fiber links. It is shown to be superior in terms of blocking probability, bandwidth blocking probability, and acceptable computational speed compared to the standard and published benchmarks at different traffic loads.
{"title":"Machine-learning-based impairment-aware dynamic RMSCA in multi-core elastic optical networks","authors":"Jaya Lakshmi Ravipudi;Maite Brandt-Pearce","doi":"10.1364/JOCN.530035","DOIUrl":"https://doi.org/10.1364/JOCN.530035","url":null,"abstract":"This paper presents a routing, modulation, spectrum, and core assignment (RMSCA) algorithm for space-division-multiplexing-based elastic optical networks (SDM-EONs) comprising multi-core links. A network state-dependent route and core selection method is proposed using a deep neural network (DNN) classifier. The DNN is trained using a metaheuristic optimization algorithm to predict lightpath suitability, considering the quality of transmission and resource availability. Physical layer impairments, including inter-core crosstalk, amplified spontaneous emission, and Kerr fiber nonlinearities, are considered, and a random forest (RF)-based link noise estimator is proposed. A feature importance selection analysis is provided for all the features considered for the DNN classifier and the RF link noise estimator. The proposed machine-learning-enabled RMSCA approach is evaluated on three network topologies, USNET, NSFNET, and COST-239 with 7-core and 12-core fiber links. It is shown to be superior in terms of blocking probability, bandwidth blocking probability, and acceptable computational speed compared to the standard and published benchmarks at different traffic loads.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"F26-F39"},"PeriodicalIF":4.0,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10675734","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169704","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Free-space optical (FSO) communication can reduce the routing complexity of data center networks (DCNs), not only ensuring high-capacity optical switching, but also owning similar flexibility to the wireless connectivity. Presently, mainstream wireless-optical switching units (WOSUs) have adopted serial beam control for unicasting. As a two-dimensional planar structure composed of meta-atoms with special electromagnetic properties arranged in a certain way, a passive metasurface (PMF) has strong parallel beam regulating capability. In this paper, we propose a wireless-optical intra-DC interconnection scheme based on PMFs. Through a real PMF chip, by adjusting the polarization of an incident beam, the power distribution between normal and abnormal reflected beams can be controlled. When both reflected beams are assigned with power, we obtain a pair of beams reflected in parallel, thus simultaneously communicating with two racks. In fact, we can perform 1-to- $N$ ( $N ge 2$ ) multicast by using cascaded PMFs, and 1-to- $N$ means that each source rack can communicate with $N$ destination racks, i.e., a wide communication coverage range. However, the cascaded PMFs may result in huge chip costs and high accumulated power loss. In this paper, we only demonstrate the 1-to-2 communication ability of our PMFs, so the communication coverage may be still limited. Hence, by introducing electrostatic drive to control the PMF rotation, each source rack can achieve a wider communication coverage range than before. As a result, all racks within the coverage range have the ability to establish communication links with the source rack. To this end, we also design a neural network to mimic the PMF rotation, further improving the communication capability between racks. We then propose a DCN-topology reconfiguration algorithm supporting the coexistence of unicasting and multicasting, in order to make real-time decisions on the matching between FSO ports with minimized latency. Finally, we established a proof-of-concept prototype system to demonstrate the parallel beam control of our PMF. The test results show that the angle error is less than 0.1°, satisfying the accuracy requirements of WOSUs, and our PMF always has a good polarization response with the insertion loss of less than 2 dB under various azimuth angles or rotation planes.
{"title":"Universal strategy study of applying passive metasurfaces to an intra-DC wireless-optical interconnection","authors":"Weijie Qiu;Weigang Hou;Jianou Huang;Xiangyu He;Pengxing Guo;Lei Guo","doi":"10.1364/JOCN.521214","DOIUrl":"https://doi.org/10.1364/JOCN.521214","url":null,"abstract":"Free-space optical (FSO) communication can reduce the routing complexity of data center networks (DCNs), not only ensuring high-capacity optical switching, but also owning similar flexibility to the wireless connectivity. Presently, mainstream wireless-optical switching units (WOSUs) have adopted serial beam control for unicasting. As a two-dimensional planar structure composed of meta-atoms with special electromagnetic properties arranged in a certain way, a passive metasurface (PMF) has strong parallel beam regulating capability. In this paper, we propose a wireless-optical intra-DC interconnection scheme based on PMFs. Through a real PMF chip, by adjusting the polarization of an incident beam, the power distribution between normal and abnormal reflected beams can be controlled. When both reflected beams are assigned with power, we obtain a pair of beams reflected in parallel, thus simultaneously communicating with two racks. In fact, we can perform 1-to-\u0000<tex>$N$</tex>\u0000 (\u0000<tex>$N ge 2$</tex>\u0000) multicast by using cascaded PMFs, and 1-to-\u0000<tex>$N$</tex>\u0000 means that each source rack can communicate with \u0000<tex>$N$</tex>\u0000 destination racks, i.e., a wide communication coverage range. However, the cascaded PMFs may result in huge chip costs and high accumulated power loss. In this paper, we only demonstrate the 1-to-2 communication ability of our PMFs, so the communication coverage may be still limited. Hence, by introducing electrostatic drive to control the PMF rotation, each source rack can achieve a wider communication coverage range than before. As a result, all racks within the coverage range have the ability to establish communication links with the source rack. To this end, we also design a neural network to mimic the PMF rotation, further improving the communication capability between racks. We then propose a DCN-topology reconfiguration algorithm supporting the coexistence of unicasting and multicasting, in order to make real-time decisions on the matching between FSO ports with minimized latency. Finally, we established a proof-of-concept prototype system to demonstrate the parallel beam control of our PMF. The test results show that the angle error is less than 0.1°, satisfying the accuracy requirements of WOSUs, and our PMF always has a good polarization response with the insertion loss of less than 2 dB under various azimuth angles or rotation planes.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"929-940"},"PeriodicalIF":4.0,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yang Zhao;Yi Lin;Yunbo Li;Dechao Zhang;Yucong Liu;Yu Zheng;Dong Wang;Sheng Liu;Shan Cao;Haoyu Feng;Han Li;Xiang Liu
With the development of service, there are demands for flexible connection scheduling based on customer-side services, fast connection response, and efficient scheduling of line-side resources in an optical transport network (OTN). This article proposes a service-oriented multi-layer resource scheduling architecture and algorithm based on the introduction of OTN fine grain technology. Specifically, we propose a service awareness scheme that enables OTNs to better match service requirements. Based on a centralized distributed collaborative architecture, we extend a path computation element communication protocol (PCEP) to achieve more efficient connection distribution. We further design a resource scheduling strategy and an algorithm based on multi-layer collaboration to perform differentiated scheduling for services with different priorities. Through the mechanisms and algorithms proposed, it is technically feasible to achieve precise matching of customer-side services, flexible scheduling of connections and resources, and efficient service response, and promote the evolution of the OTN service supply from traditional fixed point-to-point connections to flexible service-oriented optical network (SOON) connections.
{"title":"Multi-layer resource scheduling architecture and algorithm for a service-oriented optical network based on a fine grain OTN","authors":"Yang Zhao;Yi Lin;Yunbo Li;Dechao Zhang;Yucong Liu;Yu Zheng;Dong Wang;Sheng Liu;Shan Cao;Haoyu Feng;Han Li;Xiang Liu","doi":"10.1364/JOCN.527736","DOIUrl":"10.1364/JOCN.527736","url":null,"abstract":"With the development of service, there are demands for flexible connection scheduling based on customer-side services, fast connection response, and efficient scheduling of line-side resources in an optical transport network (OTN). This article proposes a service-oriented multi-layer resource scheduling architecture and algorithm based on the introduction of OTN fine grain technology. Specifically, we propose a service awareness scheme that enables OTNs to better match service requirements. Based on a centralized distributed collaborative architecture, we extend a path computation element communication protocol (PCEP) to achieve more efficient connection distribution. We further design a resource scheduling strategy and an algorithm based on multi-layer collaboration to perform differentiated scheduling for services with different priorities. Through the mechanisms and algorithms proposed, it is technically feasible to achieve precise matching of customer-side services, flexible scheduling of connections and resources, and efficient service response, and promote the evolution of the OTN service supply from traditional fixed point-to-point connections to flexible service-oriented optical network (SOON) connections.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"F13-F25"},"PeriodicalIF":4.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141641239","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Space division multiplexed elastic optical networks (SDM-EONs) enhance service provisioning by offering increased fiber capacity through the use of flexible spectrum allocation, multiple spatial modes, and efficient modulations. In these networks, the problem of allocating resources for connections involves assigning routes, modulations, cores, and spectrum (RMCSA). However, the presence of intercore crosstalk (XT) between ongoing connections on adjacent cores can degrade signal transmission, necessitating proper handling during resource assignment. The use of multiple modulations in translucent optical networks presents a challenge in balancing spectrum utilization and XT accumulation. In this paper, we propose a dual-optimized RMCSA algorithm called the Capacity Loss Aware Resource Assignment Algorithm (CLARA+), which optimizes network capacity utilization to improve resource availability and network performance. A two-step machine-learning-enabled optimization is used to improve the resource allocations by balancing the tradeoff between spectrum utilization and XT accumulation with the help of feature extraction from the network. Extensive simulations demonstrate that CLARA+ significantly reduces bandwidth blocking probability and enhances resource utilization across various scenarios. We show that our strategy applied to a few algorithms from the literature improves the bandwidth blocking probability by up to three orders of magnitude. The algorithm effectively balances spectrum utilization and XT accumulation more efficiently compared to existing algorithms in the literature.
{"title":"CLARA+: dual machine learning optimized resource assignment for translucent SDM-EONs","authors":"Shrinivas Petale;Suresh Subramaniam","doi":"10.1364/JOCN.527846","DOIUrl":"10.1364/JOCN.527846","url":null,"abstract":"Space division multiplexed elastic optical networks (SDM-EONs) enhance service provisioning by offering increased fiber capacity through the use of flexible spectrum allocation, multiple spatial modes, and efficient modulations. In these networks, the problem of allocating resources for connections involves assigning routes, modulations, cores, and spectrum (RMCSA). However, the presence of intercore crosstalk (XT) between ongoing connections on adjacent cores can degrade signal transmission, necessitating proper handling during resource assignment. The use of multiple modulations in translucent optical networks presents a challenge in balancing spectrum utilization and XT accumulation. In this paper, we propose a dual-optimized RMCSA algorithm called the Capacity Loss Aware Resource Assignment Algorithm (CLARA+), which optimizes network capacity utilization to improve resource availability and network performance. A two-step machine-learning-enabled optimization is used to improve the resource allocations by balancing the tradeoff between spectrum utilization and XT accumulation with the help of feature extraction from the network. Extensive simulations demonstrate that CLARA+ significantly reduces bandwidth blocking probability and enhances resource utilization across various scenarios. We show that our strategy applied to a few algorithms from the literature improves the bandwidth blocking probability by up to three orders of magnitude. The algorithm effectively balances spectrum utilization and XT accumulation more efficiently compared to existing algorithms in the literature.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 10","pages":"F1-F12"},"PeriodicalIF":4.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141802135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 rapid development of artificial intelligence has accelerated the arrival of the era of large models. Artificial-neural-network-based large models typically have millions to billions of parameters, and their training and reasoning processes put strict requirements on hardware, especially at the chip level, in terms of interconnection bandwidth, processing speed, latency, etc. The optical network-on-chip (ONoC) is a new interconnection technology that connects IP cores through a network of optical waveguides. Due to its incomparable advantages such as low loss, high throughput, and low delay, this communication mode has gradually become the key technology to improve the efficiency of large models. At present, the ONoC has been used to reduce the interconnection complexity of neural network accelerators, where neural network models are reshaped to map into the process elements of the ONoC and communicate at high speed on chip. In this paper, we first propose a torus-based O-type mapping strategy to realize efficient mapping of neuron groups to the chip. Additionally, an array congestion information-based low-congestion arbitrator is designed and then a multi-path low-congestion routing algorithm named TMLA is presented to alleviate array congestion and disperse the routing pressure of each path. Results demonstrate that the proposed mapping and routing scheme can reduce the average network delay without additional loss when the injection rate is relatively large, which provides a valuable reference for the research of neural network acceleration.
人工智能的快速发展加速了大型模型时代的到来。基于人工神经网络的大型模型通常拥有数百万到数十亿个参数，其训练和推理过程对硬件，尤其是芯片级硬件的互联带宽、处理速度、延迟等提出了严格的要求。片上光网络（ONoC）是一种通过光波导网络连接 IP 核的新型互连技术。由于其具有低损耗、高吞吐量、低延迟等无可比拟的优势，这种通信模式已逐渐成为提高大型模型效率的关键技术。目前，ONoC 已被用于降低神经网络加速器的互连复杂度，将神经网络模型重塑后映射到 ONoC 的工艺元件中，并在芯片上进行高速通信。在本文中，我们首先提出了一种基于环的 O 型映射策略，以实现神经元群到芯片的高效映射。此外，我们还设计了一种基于阵列拥塞信息的低拥塞仲裁器，然后提出了一种名为 TMLA 的多路径低拥塞路由算法，以缓解阵列拥塞并分散各路径的路由压力。结果表明，当注入率相对较大时，所提出的映射和路由方案可以在不增加额外损耗的情况下降低平均网络延迟，为神经网络加速研究提供了有价值的参考。
{"title":"Efficient O-type mapping and routing of large-scale neural networks to torus-based ONoCs","authors":"Qiuyan Yao;Daqing Meng;Hui Yang;Nan Feng;Jie Zhang","doi":"10.1364/JOCN.525666","DOIUrl":"https://doi.org/10.1364/JOCN.525666","url":null,"abstract":"The rapid development of artificial intelligence has accelerated the arrival of the era of large models. Artificial-neural-network-based large models typically have millions to billions of parameters, and their training and reasoning processes put strict requirements on hardware, especially at the chip level, in terms of interconnection bandwidth, processing speed, latency, etc. The optical network-on-chip (ONoC) is a new interconnection technology that connects IP cores through a network of optical waveguides. Due to its incomparable advantages such as low loss, high throughput, and low delay, this communication mode has gradually become the key technology to improve the efficiency of large models. At present, the ONoC has been used to reduce the interconnection complexity of neural network accelerators, where neural network models are reshaped to map into the process elements of the ONoC and communicate at high speed on chip. In this paper, we first propose a torus-based O-type mapping strategy to realize efficient mapping of neuron groups to the chip. Additionally, an array congestion information-based low-congestion arbitrator is designed and then a multi-path low-congestion routing algorithm named TMLA is presented to alleviate array congestion and disperse the routing pressure of each path. Results demonstrate that the proposed mapping and routing scheme can reduce the average network delay without additional loss when the injection rate is relatively large, which provides a valuable reference for the research of neural network acceleration.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 9","pages":"918-928"},"PeriodicalIF":4.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142077639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
As the demand for data transmission continues to grow, it is expected that the single-carrier bit rate will reach 1.4 Tb/s, necessitating a 14 Tb/s optical interface for efficient traffic transmissions. In such scenarios, a single request could occupy the entire C-band, and therefore such a large request can be transmitted without being groomed with others, eliminating the need for wavelength cross-connect (WXC). The spatial channel network (SCN) architecture has been proposed to address this issue. In SCNs, there are two types of groomed space lanes (SLs): spatial bypass SLs, which enable end-to-end transmission without the need for WXCs, and spectrally groomed SLs, equipped with WXCs and guardbands (GBs) to integrate requests from different nodes for transmission. Because of this characteristic of the SCN, the traditional first fit algorithm cannot allocate SLs efficiently for these two distinct SLs. In this paper, we propose a dynamic routing, spatial channel, and spectrum assignment (RSCSA) algorithm that employs a granularity switching threshold to differentiate incoming requests with different SLs. The proposed algorithm allocates smaller requests to spectrally groomed SLs and larger requests to spatial bypass SLs. This approach not only maintains network flexibility but also ensures transmission efficiency. We have identified by simulation the most suitable granularity switching threshold for given networks and request matrices.
{"title":"Dynamic routing, spatial channel, and spectrum assignment in spatial channel networks based on a granularity switching threshold","authors":"Yu Zheng;Weichang Zheng;Mingcong Yang;Cheng Jin;Chenxiao Zhang;Yongbing Zhang","doi":"10.1364/JOCN.523666","DOIUrl":"https://doi.org/10.1364/JOCN.523666","url":null,"abstract":"As the demand for data transmission continues to grow, it is expected that the single-carrier bit rate will reach 1.4 Tb/s, necessitating a 14 Tb/s optical interface for efficient traffic transmissions. In such scenarios, a single request could occupy the entire C-band, and therefore such a large request can be transmitted without being groomed with others, eliminating the need for wavelength cross-connect (WXC). The spatial channel network (SCN) architecture has been proposed to address this issue. In SCNs, there are two types of groomed space lanes (SLs): spatial bypass SLs, which enable end-to-end transmission without the need for WXCs, and spectrally groomed SLs, equipped with WXCs and guardbands (GBs) to integrate requests from different nodes for transmission. Because of this characteristic of the SCN, the traditional first fit algorithm cannot allocate SLs efficiently for these two distinct SLs. In this paper, we propose a dynamic routing, spatial channel, and spectrum assignment (RSCSA) algorithm that employs a granularity switching threshold to differentiate incoming requests with different SLs. The proposed algorithm allocates smaller requests to spectrally groomed SLs and larger requests to spatial bypass SLs. This approach not only maintains network flexibility but also ensures transmission efficiency. We have identified by simulation the most suitable granularity switching threshold for given networks and request matrices.","PeriodicalId":50103,"journal":{"name":"Journal of Optical Communications and Networking","volume":"16 9","pages":"905-917"},"PeriodicalIF":4.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142077619","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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：