Welcome

News

>

China Mobile’s Li Yunbo: Build a new type of all-optical network based on OXC and implement AI applications in stages

China Mobile’s Li Yunbo: Build a new type of all-optical network based on OXC and implement AI applications in stages

Facing the computing power network, new demands have been put forward for the optical transmission network in terms of architecture, bandwidth, latency, etc. The optical network needs to be transformed and upgraded to build an optical base that carries computing power.

 

At the 2024 China Optical Network Symposium held recently, Li Yunbo, the chief researcher of the Institute of Basic Network Technology of China Mobile Research Institute, introduced that for the computing power connection of the east-data-west-computation hub, China Mobile promotes the maturity of backbone long-distance 400G optical transmission and photoelectric linkage technology, builds a new generation of photoelectric linkage all-optical network based on 400G high-speed interconnection of OXC, and provides low-latency transmission, flat networking, and large-bandwidth guarantee capabilities.

 

At the same time, the application of AI in the optical network is gradually implemented in stages, building AI intelligent routing with dynamic scheduling capabilities to serve new applications of computing and networking.

 

Enhance the perception ability of dumb resources in the optical network, realize full parameter perception of the optical network, optical cable and resource planning and prediction, intelligent location and demarcation of faults and risk prediction, break through the intelligent simulation and decision-making technology of the optical network, promote the application of AI optimization scenarios, and realize the landing of AI in the main operation and maintenance scenarios of the optical network.

400G QPSK is a better solution for long-distance backbone transmission

 

In terms of 400G long-distance backbone transmission, China Mobile, in collaboration with the industry, through scheme design, theoretical analysis, and experimental verification, confirmed that 400G QPSK has a performance improvement of more than 50% compared to 16QAM-PCS. “400G QPSK is a better solution for long-distance backbone transmission,” Li Yunbo said.

 

The 2018km experiment based on G.652.D fiber shows that QPSK has an advantage of >1dB over 16QAM-PCS in terms of OSNR margin and transmission penalty, and the overall performance advantage is >2dB. In 2023, China Mobile completed the 400G QPSK 5616km real-time live network transmission, setting a world record for 400G transmission distance under the condition of reserving sufficient maintenance margin for live network G.652D fiber.

 

In addition, China Mobile studied and verified whether QPSK can reuse the existing G.652.D fiber. The results show that QPSK can follow G.652D to achieve a transmission distance of 1500km+. Therefore, G.652D can be fully utilized for the old to achieve the rapid deployment of 400G.

 

Li Yunbo introduced that the introduction of 400G QPSK requires spectral transmission of 12THz in the C+L band supported by the optical layer. China Mobile, in collaboration with the industry, through new doping + doping process + optimized power, the noise figure (NF) of L6T OA has been reduced from 8.5dB to below 6.5dB. In addition, for the power transfer problem brought by the stimulated Raman scattering effect (SRS), China Mobile focused onconcentrate on solving C+L key devices and led the industry to develop broadband amplifiers to solve the problem of complex operation and maintenance caused by power transfer after band expansion.

The new all-optical network of photoelectric linkage based on OXC

 

Li Yunbo pointed out that based on 400G high-speed transmission, the flexible scheduling capability of network nodes needs to be increased. China Mobile proposed a new type of all-optical network of photoelectric linkage based on OXC. The photoelectric linkage focuses on the intercommunication between the optical layer and the electrical layer services. The electrical cross-matrix is used to complete the aggregation and scheduling of small granular services, and the optical cross is used to complete the wavelength-level service scheduling. The photoelectric OAM mechanism is pulled through to realize the photoelectric networking linkage to support large-scale networking and flexible scheduling.

 

Through the centralized control system for overall networking strategy control, each node throughout the process can obtain real-time information including the topology and performance information of each wavelength service through the optical layer OAM, which is used for cross-layer service path calculation, link configuration, protection and restoration, etc.

 

between the optical and electrical layers, achieving that the original two separate networks of optical and electrical become one integrated network of optical and electrical fusion. Through the optical layer OAM, the collaborative routing and service activation of the optical and electrical layers can be realized. Each wavelength carries label information. Through OAM detection, the wavelength routing table can be dynamically obtained. Based on the topology of the optical and electrical layers, the global optimal routing is designed; by reading the performance (optical power, OSNR) of a single wavelength through the optical layer OAM, based on the OSNR calculation of reachability and wavelength availability, the electrical relay or wavelength conversion is automatically configured; at the same time, by pulling through the OAM information of the optical and electrical layers, the fault collaborative location of the optical and electrical layers can be carried out.

Implement the application of AI in optical networks in stages

 

“Facing the development of the computing power network, the scale of the optical network is constantly expanding, and it is necessary to gradually introduce AI to enhance the network service capability,” Li Yunbo said. The data analysis and information extraction capabilities inherent in AI itself make it possible to diagnose network quality, optimize business performance, reduce operational burden, and improve user experience through the collection, analysis, prediction, and decision-making of massive network data.


With the continuous development of network intelligence, China Mobile aims to achieve L4 autonomous driving in 2025. The management and control architecture of the optical network for computing power is also divided into four levels, from bottom to top, which are network element management, network management, business management, and business management. At the same time, intelligent analysis and decision-making are carried out in all aspects such as network planning, maintenance, optimization, and operation to achieve the combination of automation and intelligence.


Li Yunbo introduced that the introduction of artificial intelligence in the optical network is divided into the following aspects:


The first is to build AI intelligent routing based on the pooling of optical network resources to serve new applications of computing and networking. In the optical network with resource pooling and flexible scheduling capabilities, use AI algorithms to learn the network SLA situation (optical cable availability, latency, power consumption, resource occupation, etc.) and identify business characteristics, accurately match user demands, and realize on-demand scheduling and routing of computing and networking and configuration of bandwidth.


The second is to build wavelength-level dynamic scheduling capabilities based on the digitization of the optical layer. The OTN system integrates rich optical detection technologies. Digitization + AI enables the digital optical base to support on-demand dynamic wavelength-level dismantling and construction, multi-level network intelligent fault analysis, and continuously improves the level of intelligent operation and maintenance.


The third is the intelligent management and control of the optical network to support the global visibility of network resources. For the optical network facing the computing power network, the optical network management and control system is required to enhance the multi-dimensional evaluation of transmission capabilities, the full-scenario automatic opening, and the intelligent guarantee capabilities of business SLA through automation and intelligent technologies, to achieve accurate planning and construction of computing and networking, rapid business opening, and high-quality business guarantee.


In addition, based on the digital optical layer capabilities of optical network equipment and the AI reasoning function of the management and control platform, an optical network health prediction system can be constructed. Through AI prediction algorithms, the health of optical fibers and wave channels can be analyzed, and based on the changing trend of optical performance, the risk of deterioration-type faults can be predicted in advance, and fiber faults can be automatically located and demarcated, and the quality of splicing can be automatically analyzed. “That is, to take precautions and eliminate major network faults as much as possible in the bud stage,” Li Yunbo said.