Collaborative Layout Computing Power One Web Country | Computing | Computing Power
Researchers from the National Supercomputing Chengdu Center are inspecting equipment in the computer room. Photo by Xinhua News Agency reporter Liu Kun
In February 2022, the National Development and Reform Commission and four other ministries jointly issued a notice agreeing to launch the construction of national computing power hub nodes in eight regions, and planned 10 national data center clusters, officially launching the "East Digital and West Computing" project.
The construction of a national integrated big data center system through the "East Calculation and West Calculation" project is conducive to improving the overall computing power level of the country, expanding effective investment, and promoting regional coordinated development. This is not only a key measure to promote the high-quality development of China's digital economy and an effective support for achieving the "dual carbon" goals, but also a pioneering exploration of building a unified national market in the field of new infrastructure.
Ranked second in scale globally
The server installation work has been completed in the computer room of the National Integrated Computing Power Network National Hub Center located in Gui'an New Area, Guizhou Province. The on-site staff are busy debugging the equipment. It is reported that the total investment of the project is about 1 billion yuan, and the main computing power base includes three types of computing power: general computing, intelligent computing, and supercomputing. It is planned to be completed by the end of 2026, and the first phase of power on operation will be completed in the first quarter of this year.
"After years of construction, China's computing power scale has grown rapidly. A tiered optimized computing power supply system has been preliminarily established, ranking second in the world in terms of computing power scale, with an annual growth rate of nearly 30% in the past five years." Zhao Zhiguo, Chief Engineer of the Ministry of Industry and Information Technology, said that the construction of an industrial pattern with a relatively complete system, a large scale, and active innovation is accelerating. Currently, the computing industry in China accounts for about 20% of the electronic information manufacturing industry, with over 2300 enterprises above designated size.
On the one hand, the market share of the whole machine continues to rise. In the field of general computing, IDC data shows that Inspur, Xinhua San, Dell, Lenovo, and Huawei rank among the top five in China's server market, with a total market share of nearly 75% for domestic brands. In the field of intelligent computing, Inspur, Security Police, and Huawei rank among the top three in China's artificial intelligence server market, with domestic brands holding a market share of 85%. In the field of high-performance computing, China maintains a global lead in both the share of supercomputing systems and the total installed capacity of manufacturers. On the other hand, the industrial ecosystem is constantly improving. Domestic chips have begun to take shape, while X86, ARM, and independent architecture CPUs continue to deepen their scale application. AI chips such as Baidu and Cambrian are accelerating iteration and optimization. Domestic operating systems are gradually penetrating into industries such as finance, telecommunications, and healthcare. The computing industry ecosystem, such as Kunpeng Ecology, is gradually improving, covering key links such as underlying software and hardware, overall systems, and applications.
The recently released Digital China Development Report shows that as of the end of 2022, the total scale of data center racks in China exceeded 6.5 million standard racks, and the total scale of in use data center computing power exceeded 180 EFLOPS. The "East Calculation and West Calculation" project has entered the comprehensive construction phase from system layout. In 2022, the construction of computing power hubs in 8 countries including Beijing Tianjin Hebei entered the deepening implementation stage, with over 60 newly started data center projects and a new data center scale exceeding 1.3 million standard racks.
Wen Xiaojun, Director of the Institute of Electronic Information at CCID Research Institute, stated that since the official launch of China's "East West Computing" project, industry regulatory authorities have vigorously promoted the construction of a multi-level computing power infrastructure system with digital network collaboration, digital cloud collaboration, cloud edge collaboration, and green intelligence. The three major operators have accelerated the construction of computing power networks, data centers, and other computing infrastructure. In this context, the business structure of data centers, which mainly focuses on data storage and backup, has been disrupted, shifting towards supporting the computing power requirements of industry application scenarios.
Intelligent computing becomes the main driving force
With the continuous enhancement of China's computing power support capacity, new models and formats of application are accelerating to emerge. Among them, intelligent computing, as the most important driving force for the rapid development of the computing power industry, has reached a scale of 104EFLOPS, with a growth rate of 85%, and the proportion of computing power exceeds 50%. The White Paper on China's Computing Power Development Index shows that as of March 2022, there are over 20 artificial intelligence computing centers in operation and under construction nationwide, providing inclusive computing power to local areas and effectively supporting scientific research innovation and talent cultivation.
In March of this year, good news came from the Zhijiang Laboratory located in Hangzhou, Zhejiang Province: its participation in the "FAST Fine Characterized Active Repetitive Rapid Radio Burst" was successfully selected as one of the top ten scientific advancements in China in 2022. The study of rapid radio bursts relies on the analysis and calculation of a large amount of data. FAST patrols the vast universe day and night, producing up to 300TB of data every day. "With such a massive amount of data, it is difficult to achieve breakthroughs in a short period of time solely through manual analysis by scientists." Zhu Shiqiang, director of the Zhijiang Laboratory and chief designer of the intelligent computing digital reactor, introduced that relying on the large scientific device of the Zhijiang Laboratory's intelligent computing digital reactor, scientific researchers can deeply explore FAST observation data and detect celestial radiation phenomena such as rapid radio bursts on a large scale.
It is reported that digital reactors are also applied in fields such as materials, breeding, and pharmaceuticals, accelerating the empowerment of scientific research innovation and industrial development. "We have embarked on a path of using intelligent computing to assist national major scientific research and achieve rapid output of innovative achievements," said Zhu Shiqiang.
The reason why digital reactors can play a role in different fields is inseparable from the scheduling ability of heterogeneous computing power. "The bottom layer of a digital reactor consists of hardware devices such as chips and cabinets, while the middle layer is our independently developed 'Zhijiang Yaoguang' intelligent computing operating system, with the upper layer corresponding to specific applications. Pan Aimin, director of the Intelligent Computing Software Research Center at Zhijiang Laboratory, told reporters that the computing power running on chips from different manufacturers comes from different system architectures or instruction sets. Users often need to rewrite programs when switching platforms to perform computing tasks.". The "Zhijiang Yaoguang" operating system is like a blending agent, gathering heterogeneous computing resources on one platform, which can effectively solve the problems of repeated development and incompatibility, and reduce the threshold for users to use.
Currently, the supply of effective computing power in the market does not match the demand for digital transformation and customization in the industry, and the efficiency of computing power is at a moderate level. The issue of fragmented computing power deserves attention. "In the future, if the 'Zhijiang Yaoguang' system can be further popularized and applied, it may provide ideas for solving the above-mentioned problems," said Pan Aimin.
Promoting green and low-carbon development
While various industries enjoy the high value brought by computing power, they cannot ignore the issue of high energy consumption behind computing power. This is not only a challenge for energy consumption, environment, and climate, but also an unavoidable topic for the high-quality development of the computing power industry.
The recently released "2023 Research Report on the Development of Green Computing Power in China" proposes that promoting the development of green computing power will further reduce computing power costs. By building a public computing power service platform, strengthening the synergy and linkage between data, computing power, and economy, it will effectively empower the transformation and upgrading of green industries in various industries.
In recent years, green computing power has achieved positive results in policy top-level planning, technological product innovation, and empowering traditional industries to achieve low-carbon transformation. As of the end of 2022, China has built a total of 153 national green data centers, and the average designed energy utilization efficiency of large-scale or above data centers under construction nationwide has decreased to 1.3.
In March 2022, the first large-scale green zero carbon data center in China, the Three Gorges Dongyue Temple Data Center, was built and put into use on the right bank of the Three Gorges Dam Area in Yichang City, Hubei Province. In this project, Huawei adopts the industry's most advanced intelligent cooling solution to control PUE within 1.25. The intelligent lithium battery backup solution not only ensures the long-term reliable operation of the data center, but also saves 40% of the power supply and distribution area. In addition, the overall system solution, combined with an end-to-end prefabricated delivery model, helps customers achieve rapid deployment, reducing the delivery cycle by 50% compared to traditional data center construction models.
ZTE provides solutions based on different liquid cooling technologies and various localized customized solutions such as CPU liquid cooling, CPU+memory module liquid cooling, CPU+memory module+GPU liquid cooling, etc. according to different application scenarios, so that customers can choose according to their needs and costs, and help build a green data center. Among them, ZTE's G5 series server new products support liquid cooling technology and adopt cold plate liquid cooling, which can reduce the PUE of data centers to 1.15. The entire pipeline is intelligently monitored, and liquid leakage alarms are triggered in seconds, ensuring product reliability while achieving green and low-carbon development.
At present, the development of green computing power in China is in its infancy. Although various regions have launched initiatives for the development of green computing power, there is still a lack of a unified evaluation method for green computing power. Experts suggest that the advantages of institutional mechanisms should be deepened, and multiple departments should jointly improve the development mechanism of green computing power, coordinate the design of an integrated development path for green computing power, and study and formulate a green computing power development index.
In addition, the green computing power standard system is still not perfect, and attention should be paid to green standards that run through the entire lifecycle of computing power. New technology research and development of green computing power should be synchronized from the production, supply, management, and empowerment levels. We need to accelerate the development of safe, reliable, and internationally advanced computing power general technical standards in all aspects, with standards as the guide, use to promote construction, and assist in the iteration and upgrading of green computing power products. At the same time, we will guide the active application of advanced and applicable green data center technologies in computing power infrastructure, provide appropriate subsidies to enterprises purchasing domestically produced computing power products, and form a good internal circulation ecosystem.
