With the promulgation of a series of national policies, green and low-carbon development of data centers has emerged as an important part of the high-quality development strategy of data centers. The green and low-carbon development of data centers improves the energy efficiency of equipment through innovative energy-saving technologies, makes full use of renewable energy to reduce carbon emissions, and then promotes the basic supporting role of data centers as a new digital infrastructure and drives society to achieve the goal of energy conservation and carbon reduction.

1. What is a "green" data center?

Green data centers aim to maximize energy efficiency and minimize environmental impact, generally including data center performance efficiency, environmental impact, resource integration, and energy coordination.

Performance efficiency is a basic requirement, and high-reliability data centers need to meet the requirements of Class A data centers in GB50174 Data Center Design Code. On the one hand, it is necessary to keep the data center running without failure, and on the other hand, the IT systems, refrigeration, lighting and electrical systems of the data center can maintain stable and efficient operation. Environmental impact mainly refers to the impact of data centers on the external environment during normal production and operation. In the building planning and construction stage, a comprehensive environmental impact assessment should be carried out on the project including air, water, soil, sound, solid waste and ecology in combination with the use function and planning positioning of the building. Resource integration mainly examines the material selection, construction technology, equipment or material recycling and recyclability and product packaging in the data center construction process. Energy coordination focuses on energy management and energy-saving measures after data center operation.

At this stage, the evaluation of green data centers mainly considers its performance efficiency, takes into account the energy overall planning after data center operation, and improves the overall energy efficiency.

Due to the "three major problems" of high heat generation element heat dissipation bottleneck, resource utilization and low energy efficiency, data centers are also very important for green energy conservation, and the energy consumption of data centers has been guided and restricted from the national to local levels.

Figure 1: Energy use in data centers

The main energy consumption of data centers is IT equipment, power distribution systems, and HVAC systems, as shown in Figure 1. A typical data center, the energy consumption is mainly composed of four major blocks: the first large block is the IT equipment system that accounts for more than 50% of the total energy consumption of the data center, including server equipment, storage equipment and network communication equipment; The second largest block is the air conditioning system, which accounts for about 38% of the total energy consumption of the data center, of which the air conditioning and refrigeration system accounts for about 25% of the total power consumption, and the air conditioning supply and return air system accounts for about 13% of the total power consumption; The third largest block is the UPS power supply and distribution system, which accounts for about 10% of the total energy consumption of the data center, of which the UPS power supply system accounts for about 5% of the total power consumption, and the UPS input power supply system accounts for about 1% of the total power consumption; The fourth block, the remaining 1% of the total power consumption of the data center belongs to the auxiliary lighting system.

From the above analysis, it can be seen that the IT equipment of the data center is the largest energy consumption, and the largest energy consumption is the computing equipment represented by the server, so it is necessary to pay attention to the energy efficiency of the server. The energy consumption of HVAC systems in data centers accounts for more than 37% of the total energy consumption of data centers, and this part is also regarded as a very potential part of data center energy saving. The energy consumption of the power supply system accounts for about 10% of the total energy consumption of the data center, mainly from the conversion of power supply and distribution systems such as transformers and UPS, followed by lighting. Therefore, the green energy saving of data centers must be made from computing equipment, such as servers, HVAC and electrical systems.

 2. How to improve the energy efficiency of data centers?

The improvement of the overall energy efficiency of the data center is inseparable from the improvement of the energy efficiency of its constituent equipment, or the improvement of the energy efficiency of the equipment can directly and effectively improve the energy efficiency of the data center, such as improving the energy efficiency of servers, reducing power supply losses, etc., but the improvement of energy efficiency at the equipment level will encounter bottlenecks, and to a certain extent, due to technical and cost limitations, its energy efficiency cannot be improved again or the cost of energy efficiency will rise sharply. After encountering difficulties in improving equipment energy efficiency, methods to improve system energy efficiency are usually adopted, such as making the server work in a high energy efficiency state through load optimization, and optimizing the standby server. Or through airflow organization analysis, so that the refrigeration and air conditioning heat dissipation are matched in time and space, because the elasticity of the energy efficiency of the system is relatively large, usually start from a small system, gradually introduce more variables, involve more and more systems, and consider management factors, which has become a method and way to improve energy efficiency from the organizational level.

3. Product energy-saving technology
3.1 Servers

The essential role of data centers is to provide computing power resources, and the largest proportion of data center energy consumption is information equipment energy consumption. Intel, Samsung, Qualcomm, Nvidia and many other semiconductor manufacturers have made efforts, and various acceleration chips are blooming. XPU (a collective term for various server processors such as CPUs, GPUs, DPUs, etc.) has become a new track for semiconductor chip manufacturers to compete, and a clear main line of competition is gradually becoming obvious - major chip companies are building their own diversified product capabilities. Among them, based on the XPU product strategy, Intel has created a variety of heterogeneous computing resources from CPU to GPU, FPGA, IPU, etc., with XPU+oneAPI as the starting point, to create a comprehensive product portfolio from cloud to end, covering CPU, GPU, IPU, FPGA and dedicated ASICs and other solutions, solving the different performance of different processors using different architectures and different instruction sets to handle different computing scenarios, and better cope with the diversification of computing. It achieves a win-win situation of providing powerful computing power and sufficient flexibility, and has advantages in terms of power consumption, reliability, and size, solving the high cost and higher heat generation caused by the use of more processors in general-purpose computing to process massive data.

The improvement of server chip computing power and the diversification of computing scenarios are crucial for data center energy saving. Intel has greatly improved the load processing speed through dedicated acceleration chips, which means that the same workload is faster. Powered by GPUs and AI, it provides higher computing density and faster computing speed. Among them, the upcoming Intel data center GPU, codenamed Arctic Sound-M (ATS-M), is capable of delivering 150 trillion operations per second (150 TOPS). The first flagship data center GPU, Ponte Vecchio, has demonstrated exceptional advantages in complex financial services applications and AI inference and training workloads. The Habana Gaudi2 processor for high-performance deep learning AI training released in May this year, and the Ponte Vecchio GPU based on the Xe HPC microarchitecture designed for high-performance computing and AI enable end users to take full advantage of the high performance and power efficiency of the processor by supporting a variety of architectures.

In the energy consumption structure of data centers, the energy consumption of refrigeration and air conditioning systems is second only to the energy consumption of information equipment, and how to reduce the cooling energy consumption of data centers has become the key to reducing PUE. With the increase in server unit power consumption, the server power that ordinary server cabinets can accommodate often exceeds 15kW, which has reached the bottleneck of air convection heat dissipation capacity in the case of existing air-cooled data centers. As a technology with stronger heat dissipation capabilities, liquid cooling technology can help higher power density. Liquid cooling technology refers to a cooling method that uses a liquid with high specific heat capacity and high heat transfer coefficient as a working fluid for heat transfer to meet the heat dissipation needs of IT equipment such as servers. That is to say, it replaces air with liquid and takes away the heat generated by CPU, memory module, chipset, expansion card and other devices during operation. Common liquid cooling technologies include cold plate type, immersion type and spray type, liquid cooling servers can accept higher cold source supply and return water temperature, maximize the use of outdoor natural cold sources, and effectively reduce the PUE of the data center to less than 1.2. Based on the new generation of Poseidon warm water cooling technology and the third-generation Intel ® Xeon ® Scalable processor, Lenovo has created the Lenovo ThinkSystem SD650 server, which uses innovative cooling technology to improve performance and reduce power consumption. The "Siyuan No. 1" high-performance computing cluster has been built at Shanghai Jiao Tong University, with a PUE as low as about 1.1, achieving 42% energy conservation and emission reduction.

Although liquid-cooled servers have greatly improved their heat dissipation efficiency, due to the poor level of standardization, the industry has not yet had unified design specifications, so the reliability still needs to be improved, and the cost needs to be reduced. Intel is committed to proposing the requirements related to the design of cold plate liquid cooling systems, as well as the specification requirements that need to be complied with in the future liquid cooling design, providing a path and reference for the design and research of liquid cooling solutions in data centers, and jointly improving the quality of key components of cold plate liquid cooling technology through industry partners, and jointly promoting its standardization. Reduce design and use costs, thereby promoting the establishment and improvement of the cold plate liquid cooling ecosystem and promoting the maturity of the entire industry.

3.2 Cabinet-level energy-saving optimization technology

The cabinet-level energy-saving technology mainly includes: automatic adjustment of power according to the load changes caused by business volume and service type; adopt perfect power supply and distribution methods to improve the power density of a single cabinet; Technical measures such as improving computing density, unifying power supply and shared heat dissipation management.

According to Intel's measured data, using rack backup batteries to eliminate unplanned peak power consumption can increase the server shelf rate by 20%~30%. Increasing computing density can increase power efficiency by 2% and improve cabinet space utilization by increasing bus voltage to meet the needs of high-power racks. The modular design implements the concept of green energy conservation to achieve unified power supply and shared heat dissipation management. Liquid-cooled cold plate heat dissipation, with professional refrigerant, covering CPU, memory and other major components, the PUE of the whole machine is as low as 1.1, and the total cost of ownership can be greatly reduced. Processor power control tuning, automatic power adjustment according to service load, and Intel-based open processor microcode to adjust motherboard and processor voltage. Practice shows that this solution significantly improves the power density, can support up to 20kW power density of a single cabinet, and achieve an effective increase in computing power, even if pure air cooling is used, the PUE can be kept below 1.2 internally.

In addition, Intel and China Telecom jointly promote the deployment of AI energy-saving technology, with an energy-saving rate of more than 23% of the refrigeration system, and the average PUE of the computer room has been reduced from 1.49 to 1.38. At the same time, it has also cooperated with xFusion to jointly create a simplified architecture and high computing power density whole cabinet server, so that the cooling PUE is less than 1.1, and a single cabinet supports 144 high-power CPUs.

4. System energy efficiency improvement technology

Reasonable end-of-line air conditioning airflow organization is the basis and premise of energy saving and consumption reduction of air conditioning system. Increasing the return air temperature of precision air conditioners and the temperature of supply and return water of cold sources has always been the consensus of energy conservation in HVAC systems in data centers, but often the above energy-saving measures cannot meet the theoretical goal of high efficiency and energy saving due to unreasonable terminal airflow. The reasonable airflow organization is reflected in the uniform distribution of air flow in the cabinet inlet area of the data center computer room, and the vertical and horizontal temperature fields are in a relatively balanced state. The measurement and evaluation of airflow organization can be carried out with the help of relevant air volume, wind speed, and temperature field testing instruments, and the abstract airflow field can be parameterized and concretized. Through the application of measurement technology, the coupling correlation between the airflow organization and the temperature field distribution in the air supply area of the data center cabinet is explored, and on the basis of improving the airflow organization management, the terminal precision air conditioning operation mode is studied. Finally, the airflow organization and air conditioning energy saving management strategy AI is intelligent to ensure the precise control of measurement, tuning, management and energy saving.

The application of AI intelligent control technology helps improve the energy efficiency of data centers, mainly used in power supply and distribution systems and refrigeration and air conditioning systems. Increasing the use of renewable energy is a key step in reducing carbon emissions. Intel has developed a solution that integrates into existing energy grid infrastructure to create a more intelligent grid that can adapt to changing energy demand and energy sources. Intel has joined forces with some of the world's largest utility operators to form the Edge for Smart Secondary Substations Alliance to modernize grid substations and better support renewable energy. Eenedis, France's largest grid operator, recently joined the alliance and upgraded more than 800,000 secondary substations with solutions that provide real-time control of the entire network. Beijing Goldwind Huineng Technology Co., Ltd., which provides intelligent operation for wind power plants, has improved the accuracy of wind energy forecasting from 59% to 79.41% by using Intel AI solutions, CPU integration AI acceleration, and Intel ®DLBoost and Analytics Zoo.

The AI intelligent control technology is applied to the automatic control logic of the refrigeration and air conditioning system, and the control logic of the cooling source system is optimized by predicting outdoor meteorological parameters, analyzing load changes, calculating the characteristics of cold source equipment, and maximizing the use of outdoor natural cold sources to achieve high efficiency and energy saving of the refrigeration system.

High voltage direct current (HVDC) is a technology that has attracted attention, compared with traditional UPS, a DC/AC inverter link is reduced, so that the power utilization efficiency is greatly improved, compared with the traditional UPS system solution, its energy saving effect is estimated to be up to 8% higher. At the same time, due to the small number of DC transmission wires, there is no reactive power loss of inductive resistance and capacitive resistance, only the heat loss of resistance, which improves the energy-saving effect in power distribution and transmission. DC transmission requires only two wires, which saves a lot of line investment, so the cable cost is much lower.

5. Improve energy efficiency from the organizational level

The factors affecting the energy consumption of data centers involve many aspects, including server systems, air conditioning and refrigeration systems, power supply and distribution systems, computer room decoration, lighting, transformer and distribution equipment, power supply and distribution cables, etc., and even the level of operation and maintenance will affect the energy consumption of data centers. Through the sub-monitoring and analysis of the operating load and energy efficiency of each subsystem of the data center, the energy consumption problems in the operation of each subsystem are found, and the operation adjustment and technical improvement of each subsystem are carried out in a targeted manner to reduce the power consumption of the data center and improve the overall energy efficiency. Traditional data centers use manual or manual energy consumption analysis methods combined with dynamic ring systems, and it is difficult to achieve comprehensive monitoring, accurate prediction, analysis and control in the face of massive operating data and reports. Therefore, it is necessary to use data center energy consumption analysis tools to control the energy consumption of data centers at the equipment level, system level and project level.

For example, the data center energy consumption analysis tool developed by Intel mainly covers four functional sections: monitoring, analysis, prediction and control.

● Monitoring function: with network automatic discovery function, support a variety of devices and multiple protocols, support various brands of equipment, and be able to track the change trend of equipment energy consumption;

● Analysis function: timely detection of hot spots in the computer room, with server energy consumption analysis function, identify the imbalance of air conditioning and refrigeration and airflow organization, timely detection of "zombie server", with capacity analysis function, to achieve intelligent capacity management;

● Forecasting functions: capacity growth prediction, temperature health prediction, energy consumption and temperature prediction for correlated applications;

● Control functions: server energy consumption control strategy, hotspot-based cooling analysis, increase server cabinet capacity, temperature/power consumption-based application migration.

Based on the data center energy consumption analysis tool to monitor, analyze and predict the operation status and energy consumption of each equipment, system and other data information, timely find problems, and formulate accurate high-efficiency and energy-saving control strategies, rely on the control function of energy consumption analysis tools to carry out the implementation of control strategies, and realize the whole chain energy consumption management strategy of comprehensive monitoring, accurate prediction analysis and control.
Widely used in the data center industry CQC8302-2018 "Technical Specification for Data Center Infrastructure Operation and Maintenance Evaluation" has the following requirements for energy efficiency management:

1) Formulate an energy efficiency management system, clarify the collection requirements, collection methods and frequency of energy consumption data, and conduct systematic analysis through the collected data, including but not limited to:

statistically analyze the changes in the overall electricity consumption of the data center;


a. Statistically analyze the daily power consumption and average daily power consumption of the data center;

b. Statistical analysis of the composition and proportion of energy consumption in the data center;

c. Statistically analyze the changes in monthly energy efficiency indicators of data centers;

d. Statistically analyze the changes in the energy consumption of the UPS system;

e. Statistical analysis of the changes in energy consumption of air conditioning and HVAC systems;


2) Statistically analyze the changes in water consumption of the data center (air-cooled system selection is not applicable) to understand the operating characteristics of IT equipment:

a. Whether to analyze and understand the peak and valley operation period of IT equipment from the data center infrastructure to the granularity of the cabinet;

b. Whether to communicate with the relevant departments of customers or users to make predictions for the deployment of high-density IT workloads and formulate relevant response plans.

3) Ability to manage airflow organization, including but not limited to:

a. All possible air leaks in the facility building should be sealed to maintain positive pressure in the facility;

b. The flow direction of airflow in the facility should be directed, all possible air leaks should be blocked, blinds should be installed on all free U positions in the cabinet, unnecessary air outlets should be closed, and the best efficiency of cold air should be ensured.

4) There is a management system and periodic correction strategy for the operation threshold.

5) Based on the comprehensive consideration of safety and operating efficiency, establish a guideline for setting operating thresholds, set monitoring and alarm thresholds, air conditioning return air temperature, etc.

6) Conduct regular energy consumption analysis meetings to continuously improve and optimize energy consumption management strategies.

Intel's data center energy consumption analysis tool can help data center operation and maintenance personnel continue to meet the requirements of the CQC8302-2018 "Data Center Infrastructure Operation and Maintenance Evaluation Technical Specification".

The energy efficiency management at the data center project level needs to be combined with the organization's energy metering system and energy management system, integrated into the data center operation and maintenance system, fully integrated with the data center energy consumption analysis tool and operation and maintenance system, and continuously seek energy-saving potential on the basis of ensuring the safe operation of the data center, and adopt various innovative technologies to improve the overall energy efficiency level.

6. Conclusion and prospect

Although the energy consumption of data centers is relatively large, the overall energy consumption level of the data center has not increased much compared with the computing power provided, mainly due to the improvement of server energy efficiency in recent years. A paper published in the journal Science on February 28, 2020, titled "Global Data Center Energy Use Continues to Slow Amid Rapid Demand Growth". The paper calculated that between 2010 and 2018, global data center demand increased by 550%, and data center energy use increased by only 6%. Therefore, data centers, which play an important role in the digital economy, still need to be vigorously developed.

While vigorously building data centers, we should pay attention to the research and application of energy-saving and low-carbon green technologies, and at the same time, energy-saving and low-carbon technologies need to be organically integrated with management to jointly drive the efficient energy saving of data centers. "Three points of construction, seven points of management" has been the consensus of the data center industry, through the application of new energy-saving technologies such as liquid cooling technology, efficient cooling technology, and efficient chip servers, with refined operation and maintenance management, to finally achieve green energy conservation in data centers.

The application of intelligent technology promotes the green and high-quality development of data centers. AI intelligent control technology can effectively improve server computing efficiency, power supply and distribution efficiency, and cooling energy efficiency, and promote the improvement of data center energy efficiency. With the help of intelligent energy consumption analysis tools, through the implementation of module functions such as monitoring, analysis, prediction and control, energy consumption monitoring and control at the equipment level, subsystems and project-level energy consumption can be realized, helping the green and efficient development of data centers.

Therefore, the green, energy-saving and low-carbon development of data centers cannot simply pursue extremely low power use efficiency (PUE), but needs to be comprehensively considered in combination with computing power, IT equipment energy consumption, HVAC energy consumption, power supply and distribution energy consumption and other aspects.

The green and high-quality development of data centers has a leading and exemplary role at the social level. It will not only enable high-energy industries to rapidly change their development mode, accelerate the digital transformation process of thousands of industries, and achieve rapid energy conservation and carbon reduction, but also inspire other industries with a series of advanced technologies and ideas adopted in data centers to achieve the dual carbon goal. Through technology migration, these ideas can help other energy-intensive industries, including agriculture, logistics, mining, and manufacturing, reduce carbon emissions, so that these industries can accelerate the pace of transformation to carbon neutrality while realizing digital transformation, so as to achieve green and high-quality development of data centers and exert greater social value.