Building new infrastructure has become a national strategy. As a basic platform to carry storage, transmission and arithmetic needs, as well as a physical base to support new generation digital technology applications such as artificial intelligence, cloud computing and blockchain, data centers have become an important element of new infrastructure construction. With the continuous increase of relevant policies and the accelerated development of digital economy, data centers are ushering in unprecedented development opportunities. In the next three to five years, China's data centers will show the "four highs" development trend of "high concentration, high computing power, high energy efficiency and high technology".
High Clustering
Data Center Enters Cluster Development Stage
At present, the industry generally agrees that the future data center layout will increasingly show a pattern of "hot data clustering in first-tier cities and surrounding areas, and cold data clustering in resource-rich areas in the west". There are two main factors that drive this "dumbbell" clustering development pattern.
First, the power of the market. The first-tier cities represented by "North, Guangzhou and Shenzhen" converge most of the business traffic of cloud computing, Internet and finance, and the mobile payment, network transaction, online game, financial settlement, data hot backup, virtual machine migration, etc. in these businesses have high requirements on latency. Therefore, major customers of data centers (Internet companies, cloud computing vendors, financial companies, etc.) are willing to pay high premiums to store their data in data centers in first-tier cities. However, with the tight land, electricity resources and energy consumption targets in first-tier cities, large Internet companies and third-party data center enterprises are gradually laying out the peripheral areas of first-tier cities to form regional data center clusters serving the core cities.
Meanwhile, in recent years, under the guidance of some national and local policies, enterprises are increasingly locating data centers in central and western regions with sufficient power supply and suitable climate to meet their disaster recovery needs at low cost due to economic considerations.
In the future, with the improvement of network conditions, more business data with medium and high latency requirements (such as web browsing, big data processing, data offline analysis, artificial intelligence model training, etc.) can be deployed to data center clusters in the central and western regions.
Second, the policy is driving. Currently, China has entered the stage of rapid development of digital economy, and data center is an important physical base to support the development of digital economy. By coordinating the layout of data centers to promote the cross-domain circulation of data elements in the east, middle and west, driving a large number of internet enterprises and ICT enterprises from the east coast to extend their industrial chain links to the west, helping the central and western regions to develop digital economy and realize industrial upgrading, is an important direction supported by the central policy and local policies in the central and western regions.
Although some central and western regions have made a lot of attempts in introducing preferential policies to attract data centers to land, the actual effect is not ideal due to factors such as electricity cost, power grid support, network support, industry support and talents.
Under the impetus of the "new infrastructure", in December 2020 and May 2021, four ministries and commissions, including the National Development and Reform Commission, the Central Internet Information Office, the Ministry of Industry and Information Technology, and the National Energy Administration, jointly issued the "Guidance on Accelerating the Construction of a National Integrated Big Data Center Collaborative Innovation System" and the "National Integrated Big Data Center Collaborative Innovation System Implementation Plan". System Arithmetic Hub Implementation Plan" (hereinafter referred to as the documents), clearly proposing to guide the clustering of data centers to resource-rich areas in the west as well as areas around first-tier cities at appropriate distances through the layout of national data center hub nodes. The documents point out that the state will take multiple measures to guide data centers to optimize their layout and achieve clustering development by clarifying the access criteria for cluster construction, adjusting and optimizing the network structure, and strengthening the supporting guarantees in water, electricity and energy consumption indicators.
Both market and policy forces will push China's data centers to gradually enter the stage of cluster and scale development. Of course, in the long run, whether the above clustering development pattern can be realized still depends on the actual implementation of policies and the development of market demand.
High Computing Power
Data center to "high computing power center" evolution
At present, with the gradual implementation of artificial intelligence application scenarios and the accelerated pace of intelligent transformation in various industries, the whole society has put forward higher requirements on the computing and storage capacity of data centers. In line with this trend, data centers will gradually evolve to higher computing power and higher computing efficiency in the future.
First, the scale and density of single data centers will still gradually increase. Stacking more servers in a single data center has been an important way to expand computing power. In recent years, the development of intensive and large-scale data centers has accelerated. The "National Data Center Application Development Guidelines (2020)" issued by the Department of Communication Development of the Ministry of Industry and Information Technology shows that by the end of 2019, the total scale of data center racks in use in China reached 3.145 million racks, including about 1.179 million racks in super-large data centers and 1.194 million racks in large data centers, with a year-on-year scale growth rate of 41.7%. Large data centers have increased the density and number of racks and servers, driving continuous innovation breakthroughs in all aspects of data center layout, power supply, power consumption, cooling, network and operation and maintenance management technologies.
Second, heterogeneous arithmetic power is fully integrated. Arithmetic power is the foundation of AI applications, and traditional data centers relying only on CPU general-purpose servers can no longer meet the arithmetic power required by emerging application scenarios such as machine learning, unmanned driving, industrial simulation, and AI model training. In addition to deploying more 48-core or 64-core high core CPUs in data centers to cope with the surge in computing power demand, it has become an inevitable trend to introduce GPUs, FPGAs, ASICs and other heterogeneous computing power to take up the new computing power demand. To adapt to the convergence of heterogeneous computing power, related computing power virtualization scheduling, AI scheduling-oriented data center network, AI and big data hybrid computing acceleration and other technologies are rapidly developing to effectively achieve efficient utilization and flexible sharing of high-performance computing power distribution on the basis of data centers.
In addition, in order to achieve higher computing efficiency, accelerating the deep integration of computing, storage and network, adopting higher process chips, and improving the virtualization level of data center IT facilities will become the evolutionary direction of data centers.
High Energy Efficiency
Green energy saving is an important proposition in the new stage
Data center is a major energy consumer in information infrastructure. According to statistics, by the end of 2020, China's data center power consumption accounts for about 1% of the total social power consumption, with an average PUE of 1.6~1.7. The overall energy efficiency still has a gap compared with the international advanced level, and the utilization rate of renewable energy is low. According to the annual growth rate of 30% in the number of data center racks, it is expected that by the end of the 14th Five-Year Plan, the electricity consumption of data centers will account for more than 3% of the total electricity consumption of society. This poses a huge challenge to data center operations and environmental protection. Improving data center energy efficiency is not only an important means for enterprises to reduce costs and increase efficiency, but also an important path to achieve the goal of carbon neutrality.
At the policy level, we will continue to promote energy-saving renovation and promote the use of renewable energy. First, the energy efficiency standards of newly built data centers will be improved, and the "Guidance on Accelerating the Construction of a National Integrated Large Data Center Collaborative Innovation System" jointly issued by four ministries at the end of 2020 clearly requires that the PUE of large and super-large data centers nationwide be reduced to below 1.3 by 2025.
Second, the data center energy efficiency standard system and monitoring system should be improved. the Implementation Plan for the Arithmetic Hub of the National Integrated Big Data Center Collaborative Innovation System issued in May 2021 proposes to improve the comprehensive energy efficiency evaluation standard system of data centers, including not only the traditional electric power usage efficiency standards, but also indicators such as arithmetic power usage efficiency and renewable energy utilization, so as to promote the energy efficiency of data centers from more dimensions. The system will promote the energy efficiency of data centers from more dimensions. At present, some places have started to promote real-time data center energy consumption monitoring system.
Third, it is promoting the shutdown and transformation of stock data centers. As a global data center powerhouse, the U.S. government has launched a data center consolidation plan and optimization plan to shut down a number of inefficient and old data centers. The Chinese government advocates a similar approach. The Implementation Plan for the Coordinated Development of Data Centers in Beijing (2021-2023) has clearly proposed to promote the optimization of data center stock and the forward-looking layout of new data centers in accordance with the general idea of "four batches".
Fourth, the use of renewable energy will be promoted by improving mechanisms, such as expanding the scope of renewable energy market transactions and supporting supporting renewable energy power plants.
At the technical level, new green and low-carbon technologies will gradually expand their use. In terms of IT equipment, technologies such as cloud computing pool and virtualize data center resources, significantly increasing the utilization of servers at the same energy consumption level. In terms of cooling, new technologies such as indirect evaporative cooling and liquid cooling have gradually matured and can be deployed on a large scale in data centers as needed. In terms of power supply and distribution system, the effective application of technologies such as modular UPS, high-voltage DC power supply, and direct utility power supply will reduce the power loss of the power supply and distribution system. In addition, optimized operation control through artificial intelligence technology is becoming an important way to reduce data center energy consumption. At the infrastructure level, based on historical data, through neural network algorithms, data center temperature, humidity and other data are collected and analyzed to automatically optimize the adjustment of cooling systems, power supply and distribution systems, etc., which can further reduce PUE and break through the traditional energy-saving "ceiling".
High Technology
Data Center Technology Innovation Wave is Coming
Traditionally, data centers mainly provide basic resource services such as cabinets and bandwidth leasing to the market, and the core of their operators' competition is the acquisition of resources such as land, electricity, energy consumption, bandwidth and the maintenance capability of infrastructure. With the acceleration of digital transformation in the industry in recent years, especially after the new crown pneumonia epidemic, more and more enterprise customers are migrating to the cloud, which, coupled with the widespread application of AI technology, has had a significant impact on data centers.
First, data center network technology has become the focus of innovation. First, in terms of internal data center network, driven by the development of cloud computing and AI, the business represented by online search and parallel computing is rapidly taking off, requiring the support of a large number of distributed computing clusters in the data center, leading to rapid growth in the demand for internal data center stuff traffic. Solving the bottlenecks of data centers in terms of network and ensuring that data can be transferred between servers faster and more efficiently will gradually become the key to improving data center performance. Under this trend, new technologies such as lossless network, intelligent NIC, and programmable network have emerged as new hot spots and directions for technology development.
Secondly, in terms of data center interconnection, since most enterprises adopt the form of mixed public and private clouds, reducing the network latency when connecting different clouds is one of the technical difficulties.
In terms of cloud computing IaaS layer services, the current network virtualization technology is relatively lagging behind, and at the same time, the security protection of east-west traffic should be considered, which requires higher requirements for multi-data center network interconnection technology. To a large extent, this will also promote the change of China's Internet backbone network architecture.
Second, intelligent management has become an important development direction. With the increase in the proportion of large and mega data centers, the massive equipment and complex systems bring huge challenges to the efficient management of data centers. The construction of Data Center Infrastructure Management System (DCIM) can centralize the monitoring and management of key equipment in data centers, and on this basis, artificial intelligence and big data technology can be used to improve the real-time and accuracy of data collection, and research and train data models for energy-saving optimization, equipment warning and fault prediction, so as to enhance the automation and intelligence of operation and maintenance management. level. At the same time, AI with automated inspection robots can realize unattended operation and maintenance, further improving the operation and maintenance efficiency and security level of data centers.
In addition, the evolutionary innovation of upper-layer ICT equipment will continue to force data center infrastructure technology to accelerate changes, such as promoting prefabrication technology to accelerate the development to depth, and continuous optimization of power supply and distribution system and cooling system to adapt to new business trends.
The competition of data center operators will no longer be limited to the ability to acquire resources, but the ability to innovate technology and apply new technologies will gradually become the new core competitiveness.
Summary
In the era of digital economy, with the widespread application of new generation digital technologies such as cloud computing, artificial intelligence and blockchain, the status of data center is moving from the edge to the core of information infrastructure and becoming an important base for carrying and leading technological innovation. The next 3~5 years will still be a period of rapid development for data centers. We must clarify the development trend of data centers in accordance with the requirements of economic and social development, and jointly promote the high-quality development of data centers.
