This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy
Computing power has become a key driver of socioeconomic development, as computing networks are the infrastructure of the digital economy. China Mobile Zhejiang recently launched its F5G-A 1 ms AI-centric optical network, which will both enable the development of AI services in Zhejiang and stimulate the regional economy.
The digital economy drives computing power demand, and policies guide development directions
AI applications are currently driving demand for computing power in many industries. Computing power is critical for processing massive amounts of data and enabling complex model training, especially in sectors like smart manufacturing, fintech, digital entertainment, and smart city governance. Video interaction services like cloud eSports and cloud rendering also require high network quality, as they rely on network latencies under 2 ms. However, the recent increase in foundation model training has widened disparities in computing power availability in many different regions and industries. As such, computing resources need to be integrated and optimized promptly.
Against this backdrop, policymakers in China have prioritized the construction of computing infrastructure. In October 2023, the Chinese Ministry of Industry and Information Technology (MIIT) and five other ministries jointly issued an action plan for the development of computing infrastructure. The plan proposes development quantification indicators for computing power, transport capacity, storage capacity, and application enablement, and outlines a development path for city computing networks.
Guided by these policies, China Mobile has proposed the construction of a computing network that is "centered on computing, based on networks, and powered by smart technology". This network will need to integrate numerous information technologies to provide integrated services, so China Mobile's computing network construction plan has three phases to deliver democratized computing power that can be accessed with one click, used out-of-the-box, and scheduled on demand. Phase 1 will focus on ubiquitous collaboration, Phase 2 on convergence, and Phase 3 on integration.
China Mobile has already implemented a "4 + N + 31 + X" layout for this network and is planning an "N + X" intelligent computing center layout system. Under this system, 12 intelligent computing centers have already been built in regions and provinces like Beijing-Tianjin-Hebei, the Yangtze River Delta, the Guangdong-Hong Kong-Macao Greater Bay Area, Chengdu-Chongqing, Inner Mongolia, Ningxia, Guizhou, Heilongjiang, Shandong, and Hubei. These computing centers already provide 11 EFLOPS of power and cover all of the major hubs defined in China's national "Eastern Data, Western Computing" initiative.
In terms of network construction, China Mobile has built the integrated Jiuzhou intelligent computing network to enable the convergence of data, smart technology, and computing power, with the aim of achieving collaborative development of computing power in the eastern, central, and western parts of China. And in green development, China Mobile has proposed new industry standards for smart green computing campuses to ensure infrastructure development and service supply remain low-carbon.
A 1 ms AI-centric optical network lays a solid computing foundation for Zhejiang
In Zhejiang in particular, China Mobile has developed an F5G-A-based 1 ms AI-centric optical network that covers the entire province. The network uses intelligent computing nodes to meet specific local backbone, metropolitan, network scheduling, and reliability needs.
One computing network for the entire province
China Mobile Zhejiang first used upgraded OXC all-optical grooming to optimize network structure. Administrative region restrictions have been eliminated to build all-optical mesh networks, and one-hop connection links between backbone nodes now form a 3-dimensional all-optical network that supports flexible scheduling. This enables all-optical grooming of computing nodes and one-hop connection for all cities in the province, which creates a "highway" for data and computing transmission that can be used to distribute computing resources appropriately across the province. Before, geographical restrictions significantly hindered the efficiency of scheduling between the province's scattered computing resources. Now, computing resources in Zhejiang can flow freely and be allocated on demand across the region, just like electricity.
1 ms latency to access DC
China Mobile Zhejiang then fully upgraded its metro networks to enable flexible computing access networks that can more efficiently match computing resource supply and demand. The carrier is also continuing to move OTN to integrated service access equipment rooms to improve OTN reach and enable all-optical 10 gigabit access upgrades in key areas. This enables 1 ms latency to access compute, which allows users to enjoy faster access and on-demand computing power. This additional ultra-low latency meets the needs of latency-sensitive services like cloud eSports and cloud rendering, and can be used to power stable high-speed networks for complex computing tasks like foundation model training. All this greatly improves the efficiency of AI services and ensures that computing resources can respond to user requirements as quickly as possible.
Integrated computing network scheduling
China Mobile Zhejiang next developed an all-optical transport capacity map that integrates computing node information into the network topology to form a unified view of its computing networks. Using advanced latency evaluation technologies, the company can accurately analyze latency coverage and quickly evaluate the network latency coverage of computing nodes. Path computation is then enabled based on factors like latency, bandwidth, and availability and computing network solutions can be recommended based on user requirements. This integrated computing network scheduling ensures efficient and convenient computing resource allocation. Users can also select a computing configuration most suited to their service needs, which simplifies AI service deployment and slashes costs. Together, these improvements deliver better user experience and service flexibility.
One-stop security assurance
Finally, China Mobile Zhejiang is using ASON technology to provide multi-route protection for services on the 1 ms AI-centric optical network. If a fault occurs on the network, services are restored automatically to ensure network reliability. The private network also supports online simulation and fault analysis to identify service risks in advance. Data security and network stability are critical to AI service development, and this security assurance provides powerful protection for enterprise data and services. Enterprises can deploy services on a 1 ms AI-centric optical network without worrying about service interruption and data loss caused by network faults. This is promoting stable development within Zhejiang's AI industry.
Using the AI+ engine to power diverse industries and create better smart living
Integrating computing resources to improve AI industry collaboration efficiency
Another achievement this new network has enabled is the world's first cloud-native computing power-based grid connection, which China Mobile Zhejiang completed in partnership with the Hangzhou Municipal Development and Reform Commission, Zhejiang University, Westlake University, Turing Town, and China Mobile Cloud. The Baichuan computing platform they created connects and efficiently manages the cross-industry, cross-region, and cross-level computing resource pools of Zhejiang University's QiZhenGPT, Westlake University, Turing Town, Zhejiang Lab, and China Mobile (Hangzhou). This computing power resource coordination has helped the province develop a new computing network service capability system that enables service integration and ensures high-quality computing access and high-speed connection between China Mobile's and third-party computing centers. For example, if an enterprise in Turing Town needs to power a large-scale computing task and there are available resources at Westlake University, the enterprise can invoke those resources through the Baichuan platform.
China Mobile Zhejiang's efficient computing resource integration eliminates barriers in computing power caused by siloed organizations. It fully utilizes Zhejiang's abundant education, research, and industry resources to accelerate local R&D and application of AI technologies. This is turning the province into a frontrunner in industry AI application.
Building a multi-layer computing power layout to meet diverse service needs
China Mobile Zhejiang has also used this 1 ms AI-centric optical network to form a computing tidal network while transforming the layout of its data centers. It has set up a multi-layer computing power layout for 5 central nodes, 11 cities, and 31 counties, with more than 2 million vCPUs in general-purpose computing power. The carrier's "N + X" intelligent computing center system delivers over 4.5 EFLOPS of intelligent computing power. Its first batch of nodes for a China Mobile Intelligent Computing Center (Hangzhou) have been built.
Figure: China Mobile Zhejiang officially launching the F5G-A 1 ms AI-centric optical network
The tidal network delivers more elastic computing power and smarter algorithms to help form a foundation for Zhejiang's AI industry. The multiple layers of this computing power layout can also meet the needs of many different industry needs, as it provides the bandwidth and latency required by data-sensitive industries like finance and government for local data storage and cloud-based training, while simultaneously enabling large Internet enterprises to process massive amounts of data and perform complex model training.
Driving AI implementation to facilitate digital transformation across industries
In other sectors, China Mobile Zhejiang's 1 ms AI-centric optical network is helping traditional industries, like healthcare, services, manufacturing, and agriculture, digitally transform with the use of AI technologies. The network provides enterprises with readily accessible, on-demand computing services that significantly lower the threshold for AI use. Ningbo KFBIO, for example, uses the network's computing power to analyze and diagnose biomedical images faster, improving the efficiency and quality of its medical services. Hangzhou ustyle Technology uses the network's low latency to optimize the intelligent scheduling algorithms of its supply chain management system, which has significantly improved the accuracy and punctuality of their logistics and distribution system. Hangzhou Linker Technology uses the network's wealth of computing resources to accelerate the R&D and iteration of AI algorithms for more accurate image recognition and behavioral analysis. These examples reflect the rapid pace of regional digital economic development that is being enabled by China Mobile Zhejiang's network.
Improving computing network operation efficiency to pioneer new digital and intelligent transformation
At the core of this network is the transport capacity map jointly developed by China Mobile Hangzhou and Huawei, which is designed for intelligent and flexible scheduling of computing power. Its digital capabilities help control global network information and allow for comprehensive network planning and precise network construction. Its transport capacity overview combines a latency map, GIS map, utilization map, and availability map into one, and displays the measured electrical-layer latency. Computing node information has also been integrated into the network topology to form a unified view of computing networks, which helps planning and construction personnel ascertain network status faster, identify network resource bottlenecks, and complete bidirectional coverage evaluation from computing to network and vice versa.
The transport capacity map uses a data evaluation technology to analyze and visualize latency coverage in the form of circles, which enables faster, more intuitive evaluation of the network coverage of computing nodes measured by latencies. The map can also display nodes that can be covered by the latency circle at each level of the selected computing power pool. Latency levels can then be flexibly customized to display latency circles based on different customer needs.
Computing accessibility evaluation using real-time network awareness and multi-factor path computation can additionally help users find the optimal computing power network solutions for their latency, bandwidth, and availability needs while ensuring efficient allocation of computing network resources. All of this enables upper-layer systems, including the computing-network brain, to optimize allocation of computing network resources.
China Mobile (Hangzhou) plans to continue innovating and developing ubiquitous deterministic transport capacity to make computing services available anytime, anywhere and support the development of Zhejiang's digital economy.
Outlook: Innovative services create a bright future for Zhejiang's AI industry
The launch of the 1 ms AI-centric optical network is a milestone event for computing networks. China Mobile Zhejiang will continue to reinforce computing with network connectivity, upgrade its information infrastructure, consolidate its computing network foundation, provide high-quality digital and intelligent capabilities, and use its AI+ engine to power many different industries. The carrier hopes to stimulate the development of the regional economy in unprecedented ways. With the support of its partners and peers, China Mobile Zhejiang is committed to helping Zhejiang create a brighter future in the AI industry and promoting the development of the digital economy.
