当前位置:网站首页>Six trends and eight technologies of high-performance computing in data centers under "data center white paper 2022" and "computing from the east to the west"
Six trends and eight technologies of high-performance computing in data centers under "data center white paper 2022" and "computing from the east to the west"
2022-07-23 12:12:00 【High performance server】

Count from east to west | Data Center | High performance computing
Computing power center | Low carbon and energy saving | GPU The server

In the age of digital economy , Computing power is gradually becoming a new productivity , Widely used in all aspects of society , Provide a driving force for the digital transformation of all walks of life . The data center is the physical carrier of computing power , Is the key infrastructure for digital development .
at present , China's data center industry is evolving from high-speed development to high-quality development . In layout , National integrated data center 、 Introduction and implementation of new data center and other policy documents “ Count from east to west ” Implementation of the project , Point out the direction for the coordinated and integrated development of the data center , Promote the continuous optimization of the industrial layout of the national data center .
The fourth industrial revolution driven by digital technology has brought great reform to human production and life . As the physical basis of various digital technology applications , The industrial enabling value of data centers is gradually becoming prominent . In the world , Most countries in the world are actively guiding the development of data center industry , The data center market is expanding , Active investment and M & A, increasingly fierce competition . At home ,“ New infrastructure ” Development and “ the 14th Five-Year Plan ” Putting forward the goal of Building Digital China , Provide important guidance for the construction of digital infrastructure in China . China's data center industry has entered a new stage , The size of the data center has grown steadily , Low carbon 、 High-quality 、 The pattern of coordinated development is gradually taking shape .

Steady and rapid development in six directions
One 、 Data center spending growth : International stability , Domestic acceleration
The increase of global data centers is relatively stable . According to the annual increase of global servers ,2015 - 2021 The scale of new servers put into use in is relatively stable , The net added value is also relatively stable . It is expected that the scale of data center will continue to grow steadily in the next few years .

Annual increase of global servers ( mw )
China's data center rack scale continues to grow steadily , The scale of large and above data centers is growing rapidly . In recent years , The scale of China's data center rack is growing steadily . according to 2.5kW Statistics of standard racks , To 2021 By the end of the year, the scale of data center racks in use in China had reached 520 Ten thousand units , The average annual growth rate in the past five years has exceeded 30%. among , The rack size of large and above data centers is growing more rapidly . according to 2.5kW Standard rack statistics , The rack size is 420 ten thousand , Proportion 80%.

China's data center rack scale
Two 、 Data center market revenue : International stability , High domestic growth rate
Data center market revenue ,2021 In, the global data center market exceeded 679 Billion dollars , a 2020 Annual growth 9.8%. expect 2022 The annual market revenue will reach 746 Billion dollars , The overall growth rate remained stable .

Global data center market size
In the new infrastructure 、 Driven by national policies such as digital transformation and the long-term goal of Digital China, as well as the demand of enterprises to reduce costs and increase efficiency , China's data center business revenue continues to grow at a high speed .2021 year , The market revenue of China's data center industry has reached 1500 Billion yuan , The average annual compound growth rate in recent three years is 30.69%. With the development of various regions in China 、 The deepening of digital transformation in various industries , The market revenue of China's data center industry will maintain a continuous growth trend .

China's data center market scale
3、 ... and 、 Emerging markets are strong , There are various scenes in China
globally , Driven by the global digital economy , India 、 Emerging markets such as South Africa have gradually strengthened policy support and industrial investment in data centers , Become an important factor driving the growth of global data center demand .2015 Year initiated “ Digital India ” Provide continuous support for digital development in India , The construction of big data center is one of the important contents of the plan .
2021 year , Issued by the Ministry of communications and digital science and technology of South Africa 《 Draft national data and cloud policy 》, Guide the construction of national high-performance computing and data center . High performance computing and data centers are mainly composed of Sentech and Broadband Infraco The merger of two existing data centers , Will serve governments at all levels 、 Enterprises and universities provide cloud services . The government led data center construction will lay a solid foundation for the development of digital economy in South Africa .
Market aspect , India is increasingly rich in intelligent terminal devices and digital consumption scenarios , The demand for data storage surged . meanwhile , The aggravation of the COVID-19 has turned more offline business activities to online . The vigorous application of digital technology and consumer demand provide a good external environment for the development of India's data center industry . Multinational enterprises are optimistic about the development of India's digital economy , Strengthen digital service capacity through data center construction .2020 year , Amazon will invest in the Indian state of trengana 28 Billion dollars to build a new digital Center , This will be Amazon's second cloud computing area in India . Amazon's investment in the Indian data center market continues to increase , It will better comply with India's data localization policy .
meanwhile ,2019 In, the company announced that it would invest in the Indian market in the next three years 10 Billion dollars , Among them, the investment in data center accounts for a large proportion . Compared with the active Indian market , Multinational corporations' investment in the South African market is still in its infancy .2019 Microsoft took the lead in announcing the opening of a data center in South Africa , Become one of the world's leading cloud service providers in Africa ;2020 Amazon's first cloud data center in South Africa was officially opened in , It provides convenient cloud computing services for millions of customers . With the rapid development of digital economy in South Africa , More data centers and cloud service providers are expected to enter the South African market .
From the perspective of China , High-tech 、 Digital transformation 、 Diversified computing power demand scenarios such as end consumption are emerging , The effect of computing power is prominent .
In high technology , Highly complex computing scenarios require more high-performance computing power , Supercomputing can provide medical services with high-performance computing power 、 Aerospace 、 Provide support for exploring areas . at present ,E Level supercomputing has become an important direction for countries all over the world to compete in the field of supercomputing , There is still much room for improvement in computing power and efficiency in China . In terms of industrial digital transformation , Internet 、 signal communication 、 Modern service industries such as finance need to provide services for a large number of end customers , The application of digital technology in enterprises is relatively mature , High degree of digitization .
With the introduction of the 14th five year plan and other policies , And the continuous evolution of technology R & D and business applications , The state grid 、 Southern power grid 、 petrochina 、 Sinopec and other traditional Chinese industrial enterprises have also begun to actively promote the construction of computing infrastructure , Provide support for enterprise digital transformation . In terms of mobile consumption and intelligent terminals , In recent years , The number of mobile terminal users and intelligent terminal devices in China is growing rapidly , The application scenarios of terminal devices are constantly enriched , The demand for real-time computing power is increasing .

Heterogeneous computing power requirements in different scenarios
In computing power , The diversified development trend of China's data center is gradually emerging , Smart computing centers and edge data centers will maintain rapid growth . For a long time , China's data centers are mainly based on general computing power , Supercalculation 、 The application and number of Intelligent Computing and edge data centers need to be increased . With the development of high-performance computing in China 、AI The increasing demand for computing and edge computing , Supercalculation 、 Smart computing and edge data centers will be further developed , Computing power will show a diversified development trend .
at present , The data center of general computing power is still the main force in the market , By rack size 90% above ; Supercomputing Center is mainly used in major national scientific research fields , There are few business application scenarios ; The Intelligent Computing Center has gradually moved from early experimental exploration to commercial pilot . Although the existing scale is not high , However, with the enrichment of various artificial intelligence application scenarios in China , The demand for intelligent computing will grow rapidly , The scale growth rate is expected to reach 70%. The edge data center can provide real-time computing power for intelligent terminals and Internet of things devices . With the acceleration of China's digital transformation , Including the development of industrial Internet , The demand for edge computing will be further enhanced , The scale growth of edge data centers is expected to reach 30%.

Four 、 Big data center operators look at overseas markets
To enrich business sources , Major data center operators actively expand overseas markets , Expand industrial territory . Take China Telecom 、 China Mobile 、CenturyLink and NTT Communications Basic telecom operators represented by , As well as Equinix、 Third party data center operators represented by Wanguo data , With the first mover advantage , Has mastered large-scale data center resources in China .
Basic telecom operators and third-party data center operators are in the process of continuous development , Relying on resources 、 Capital and technology R & D advantages , By cultivating private brands 、 Broaden international marketing channels 、 Provide all-round supporting data center services , Accelerate the layout of data centers around the world .
For example, the world's leading data center operator Equinix, In Singapore 、 Hong Kong 、 Tokyo 、 Sydney 、 Jakarta and other places have data center layout ;2021 year , Wanguo will initially form the market layout of Southeast Asia , And will invest in the construction of super large-scale data center parks in Malaysia and Indonesia ; China Mobile International Co., Ltd (CMI)2021 In, the data center in Frankfurt, Germany was put into use .
As more and more giant enterprises are not limited to the local market , Actively expand the industrial layout of the data center , The competitive arena of global data centers will be further expanded .“ Sea ” Enterprises need to strengthen technology research and development and resource acquisition , To ensure greater competitive advantage .
The data center industry is ushering in a wave of mergers and acquisitions , Large transactions are frequent , The market structure may show a stronger trend . In the context of global low-carbon energy conservation strategy , The resources of developed cities and surrounding areas are becoming more and more difficult to obtain . meanwhile , Because new data center projects often need 1-2 Years of construction time , Financing is convenient 、 Sufficient funds 、 Data center market participants with flexible organization tend to participate through mergers and acquisitions “ Buy ” Data center resources or projects , To expand business scale .
Take the situation in China as an example . At present, there are two ways of M & A in the industry .
1、 Resource acquisition
More specific to the approved energy consumption quota 、 land use right 、 Workshop, etc . The purpose is to fill the shortage of data centers in some regions , Reserve scarce resources in high-quality areas .
2、 Acquire mature data center projects
Usually target the profitability of the acquired data center , Or new enterprises entering the data center industry across borders intend to speed up the layout , Shorten the construction period , Reduce the difficulty of transformation .
2021 year ,Equinix To spend 3.2 $billion acquisition of Nigerian data center and connection service provider MainOne Fill the gap in the African market ; Baoneng Chuangzhan to 16.5 Billion yuan to acquire Dr. Peng's 5 Data center 9 An aircraft building , Accelerate your entry into the data center market . M & a mode , Enterprises can quickly obtain scarce resources , Improve market share and industry position , Even accelerate the entry into the blank market that has not been involved . The expansion of basic resources will effectively promote the income growth of enterprises , After acquiring more resources, we will continue to accelerate the accumulation of resources , So the stronger, the stronger .
5、 ... and 、 Low carbon environmental protection : Technology continues to improve
The rapid emergence of innovative technologies in the data center , Continuous improvement of management and financial system , It provides an effective boost for the green and low-carbon development of the data center . In terms of technological innovation , Liquid cooling 、 Cold storage 、 High voltage DC 、 Waste heat utilization 、 Energy storage power station and other technologies, as well as solar energy 、 Utilization of renewable energy such as wind energy , Can further reduce data center energy consumption and carbon emissions . In management and Finance , The industry-leading data center promotes the green transformation of the data center by establishing a green data center management system and an internal carbon pricing system ; Establishment and improvement of green power certificate and green power trading market mechanism , It can effectively stimulate the enthusiasm of data centers to use green energy and reduce carbon emissions .

technology 、 Management and financial innovation help the green and low-carbon development of data center
Internet giants actively promote the use of green energy , Promote energy conservation and emission reduction in data centers . Google 、 Apple and Facebook Actively announce the progress of renewable energy use , And on 2017 year 、2018 Years and 2020 Implemented in its operating system 100% Renewable energy use . Microsoft and Amazon plan to 2025 Years of implementation 100% The goal of using renewable energy . In order to better promote energy conservation and emission reduction in the data center , Google in 2020 year 9 Month proposed , To 2030 In, all data centers around the world should realize hourly real-time renewable energy supply , namely 24/7 Zero carbon operation management plan . By optimizing the backup power system in the data center , Including energy storage batteries and low-carbon fuels ( Hydrogen fuel ), Microsoft is in 2020 Successfully completed the test of hydrogen fuel cell , This will eliminate the dependence on diesel generators for standby power supply , Lay a solid foundation for carbon emission reduction .
China's leading Internet companies and third-party service providers are also actively carrying out energy-saving and emission reduction technology practice , The green and low-carbon level of building and operating data centers has reached the world's leading level . Baidu cloud ( Yangquan ) Application of municipal power direct supply in Computing Center +HVDC、 Independently developed “ Zero power consumption ” Roof cooling unit and AI Optimization techniques , Average annual PUE 1.08; The capital data center of Qinhuai data ring adopts modularization 、 Green power trading, resource recovery and other technologies , Average annual PUE 1.15, Renewable energy utilization 100%. Kunshan No.1 data center of CICC data adopts energy-saving technology and clean energy procurement , Achieve carbon neutrality . Data center carbon utilization (CUE) by 0.49.
6、 ... and 、 The central and local governments work together
New data center is the direction of Industrial Development ,“ Senior four and senior three ” The connotation is constantly enriched .2021 year 7 month , Issued by the ministry of industry and information technology 《 Three year action plan for new data center development (2021-2023 year )》, Clearly point out the development direction and path of the new data center . Focus on high technology 、 High calculation power 、 High energy efficiency 、 Four characteristics of high safety , And digital cloud collaboration 、 Cloud side collaboration 、 Requirements for data network collaboration , Guide the high-quality development of China's data center . meanwhile , The action plan also starts from the layout 、 The Internet 、 Calculate the force 、 The industrial chain 、 Green low carbon 、 Security and other aspects have given comprehensive guidance to the construction of the new data center .
1、 Layout
Form a national hub node 、 Provincial Data Center 、 Echelon layout of edge data center ;
2、 The Internet
Focus on the whole country 、 Area 、 Key nodes of edge network , Promote the improvement of network quality , Promote the coordinated development of multiple networks ;
3、 Calculate the force
introduce FLOPS To assess the computing power of the data center , Guide the data center to evolve from extensive rack scale growth to high-quality computing power and efficiency improvement .
4、 The industrial chain
Encourage research and development of core technologies and standards , Strengthen the construction of industrial chain system ;
5、 Green low carbon
Accelerate the industrial application of green technology , Improve the utilization level of new energy , Optimize green management capacity building ;
6、 Security
It will improve the ability of network security guarantee and data resource management , Improve the reliability of the new data center .

Eight technologies are constantly changing
One 、 Application of new energy technology
Energy storage technology through “ Peak shaving and valley filling ” Become an important way to reduce the power cost of Data Center . The data center consumes a lot of energy , Electricity costs account for... Of the total operating costs 60%-70%. at present , Although many data centers improve the power utilization of data centers through energy-saving optimization , However, the cost of electricity still accounts for a large proportion of the total cost of Data Center . In order to balance the power consumption period of the power grid , Power supply companies usually provide peak and valley electricity prices . The data center can use the energy storage system to store electric energy in the trough period , Use during peak periods , Reduce the power cost of the data center .
Cold storage 、 Energy storage is an important solution . Cold storage can prepare cold energy when the power load is low at night , When the power load is high during the day, the prepared cold energy is applied to the air conditioning system . Energy storage is realized through energy storage devices . Lithium battery has become one of the backup energy storage schemes for the next generation data center because of its high density and high output voltage .
The integration with new energy storage technology is deepening , The energy structure of the data center has been effectively transformed , Raise the level of green and low carbon . With the tightening of data center energy consumption policy and the establishment of carbon neutrality goal , New energy power supply has gradually become an important way for data centers to achieve zero carbon emission , wind 、 light 、 water 、 The proportion of clean energy such as hydrogen will continue to increase . The data center can directly use new energy to realize energy supply , It can also indirectly promote the use of new energy through carbon emission trading .
Although new energy power generation has the advantages of clean and environmental protection , However, the supply of new energy is easily affected by natural conditions , It is difficult to guarantee its continuity . Integration with new energy storage technology , It can effectively improve the stability of new energy supply , Solve the power imbalance of renewable energy system application 、 The problem of poor stability .
Some data centers at home and abroad have begun to try large-scale power supply schemes of new energy and energy storage , It provides an important reference for the application and promotion of new energy and energy storage technology in the data center .2020 year 8 month , American energy developers Capital Dynamics With data center operators Switch Signed a power purchase agreement for solar energy and energy storage projects . Based on this Agreement ,Switch Will receive round the clock green power support . The project includes a 127MW A solar power plant and a 240MWh Battery energy storage system .
domestically ,2021 year 7 month , Century Internet's new generation Dutch storage data center project ended in Foshan smart city data center , This is a data center project to realize the application of large-scale energy storage technology . The project takes the data center as the main load object , Equipped with 2MWh Energy storage capacity ,1MW Output power . The whole energy storage system consists of energy storage containers 、PCS Warehouse and ring main cabinet , It is coupled with the photovoltaic power generation system on the AC side . Finally, the data center 10kV The high voltage side is connected to the power grid , Make the data center form a complex with variable and adjustable load , The charging and discharging strategy can be adjusted according to the demand of power grid and new energy power generation .
Two 、 The intelligence of operation and maintenance has been continuously improved
The application of intelligent operation and maintenance can effectively improve the process and standardization of data center operation and maintenance , Improve the efficiency of operation and maintenance management . With the increasing scale of Data Center , The amount of business and data carried by the data center is growing , The operation and maintenance mode of data center is also evolving rapidly . The application of intelligent operation and maintenance tools makes the operation and maintenance management of data center more efficient .
In the period of manual operation and maintenance , Data center operation and maintenance managers lack effective operation and maintenance tools , Operation and maintenance management activities depend more on their personal knowledge and technical level . With the rapid growth of the data center , The scope and difficulty of data center operation and maintenance services have been significantly improved . It is difficult to guarantee the quality of operation and maintenance services only by personal experience .
In the above case , There began to be a process 、 Standardized operation and maintenance . The purpose of process and standardized operation and maintenance is to define processes and standard documents , Promote the standardization of operation and maintenance management activities , Reduce dependence on personal knowledge and skills of operation and maintenance personnel . Automation platform operation and maintenance is a process 、 Further development of standardized operation and maintenance . Automatic operation and maintenance tools are built through the platform , Turn a large number of repetitive operation and maintenance activities into automatic operation , Improved operation and maintenance efficiency , Enhance the visualization of operation and maintenance process . Intelligent operation and maintenance is based on data , Supported by algorithms , Scene oriented , Provide intelligent solutions for data center operation and maintenance management , Realize the real-time analysis and processing of problems , Reduce alarm noise , And analyze and predict the causes of faults and problems .

Development history of data center operation and maintenance
2021 year 7 month , Ministry of industry 《 Three year action plan for new data center development (2021-2023 year )》 Clearly put forward “ For power supply and distribution 、 Refrigeration 、IT And network equipment 、 Focus on key links such as intelligent system , New data center , Build the strong , Make up for the weak ”. Policies guide the intelligent development of data center operation and maintenance management , The industry pays more and more attention to intelligent operation and maintenance .
Chinese Internet enterprises and third-party service providers actively carry out intelligent operation practice . Wanguo data Beijing No. 4 data center uses a large number of automation equipment with professional experience to replace manual operation in infrastructure construction , Improve fault alarm and response speed , Combined with independently developed operation management platform and application software , Realize efficient and high-quality operation and maintenance from local to regional and then to the whole country . Tencent Huailai Beirui Cloud Data Center , Based on the data center automation management platform developed by Tencent, Tencent Zhiwei , Built the park 、 Area 、 A three-level closed-loop management system linked by the headquarters , The second level sensitive sensing is realized by reconstructing the alarm link . Calculate through the application diagram 、 Object model and other technologies , Fast alarm convergence is realized , The accuracy of automatic fault location is as high as 99.9%.
Intelligent network operation and maintenance is one of the important application scenarios of intelligent operation and maintenance of data center . Intelligent network operation and maintenance technology can continuously promote network visibility 、 Improvement of manageability and controllability . As the size of the data center continues to expand , The complexity of the data center network is also increasing . meanwhile , Users' quality of data center network 、 Higher requirements for reliability and safety . The traditional data center network has signal interference in the process of operation and maintenance 、 Application guarantee and other issues . Poor manual operation and maintenance efficiency , When dealing with sudden network failure, it is difficult to make timely network failure treatment and operation and maintenance .
at present , The industry has proposed to “ Autopilot network ” Intelligent network operation and maintenance optimization technology represented by , The fusion AI Technology realizes the control of data center network 、 Automatic operation and maintenance and optimization , Recover and cure all kinds of network failures , Manage congestion , Support automatic network tuning and self evolution , Provide users with better network services .
stay 2021 Open data center Summit , China Academy of communications 、 Huawei 、 China Mobile Research Institute and other units jointly released 《 Super fusion data center network lossless Ethernet scenario evaluation specification 》, Slave function 、 performance 、 reliability 、 Compatibility 、 Operation and maintenance and other dimensions are important to distributed storage 、 Centralized storage and high-performance computing scenarios are evaluated . According to the evaluation results , Lossless networks are divided into five levels , Respectively, the non-destructive ability is not yet mature 、 Have basic nondestructive ability 、 Non destructive capability mature 、 Good nondestructive ability 、 Excellent nondestructive ability , Operation and maintenance is an important rating dimension .
meanwhile , Huawei in 《 Huawei data center autopilot network white paper 》 Define the autopilot network as five levels , It is pointed out that fully autonomous network is the ultimate goal of data center network development . Under the fully autonomous network , The network system itself will face any scenario , Capable of execution 、 monitor 、 The ability to analyze and make decisions , Fully realize the automation of closed-loop management , True realization “ unmanned ”.
3、 ... and 、 The construction period of the project is shortened
With the deep integration of prefabricated building technology and modular data center , The reliability and experience of the new generation of assembled modular data centers have been greatly improved . With the rapid iteration of information technology and the shortening of users' requirements for the delivery cycle of Data Center , The traditional data center construction mode is more and more difficult to meet the actual needs , Data center prefabrication has become one of the key technologies to realize the rapid construction of data center .
Data center Prefabrication Technology has been developed for many years . Early prefabricated data centers were designed in the form of all-in-one machines , A single box integrates all subsystems of the data center , It can meet the requirements of rapid deployment and emergency construction of small data centers .
Based on the all-in-one machine , The industry has gradually realized the equipment area 、 Modularization of core areas such as power distribution area , Traditional prefabricated modular data centers have emerged . Due to reliability 、 Space and standardization constraints , Traditional prefabricated modular data centers are still mainly used for small-scale and specific scenarios . With the maturity of prefabricated modular concept and the development of modular data center , The integration of prefabricated modular building technology and modular data center is deepening , A new generation of prefabricated modular data centers are beginning to appear .
Four 、 The commercial conditions of liquid cooling technology are gradually mature
Liquid cooling technology is developing rapidly , The system reliability is gradually improved . Liquid cooling technology uses liquid as heat transfer medium to exchange heat and cool near the heat source , Instead of indirect cooling through air like wind cooling . Because the liquid has a relatively high specific heat capacity , Its refrigeration effect and energy efficiency are much higher than that of air cooling , In high density 、 On a large scale 、 It has obvious advantages in data centers with high heat dissipation requirements .
According to the contact mode between the end of the liquid cooling machine room and heat sources such as servers , Liquid cooling can be divided into indirect liquid cooling technology and direct liquid cooling technology . In indirect liquid cooling , The heat source is not in direct contact with the coolant . In direct liquid cooling technology , The coolant directly contacts with the heating electronic components for heat exchange and refrigeration . Indirect liquid cooling is mainly cold plate , The single-phase cold plate liquid cooling scheme is more common . Direct liquid cooling is mainly submerged , High heat dissipation efficiency , Low noise .
at present , Liquid cooling technology has developed rapidly in China , There have been large-scale commercial cases , This is related to the continuous expansion of the scale of China's data center , The increasing power density of a single cabinet . Alibaba 、 Baidu 、 tencent 、 Huawei 、 Zhongke dawning, etc IT Companies have mature application cases of liquid cooling technology . Alibaba Renhe cloud computing data center adopts fully submerged liquid cooling server , The data center has become the first green grade in China 5A Liquid cooled data center .
With the application of liquid cooling technology , In recent years , The research on liquid cooling technology in the industry is gradually deepening .2019 year , Institute of cloud computing and big data, Chinese Academy of information and communications ( hereinafter referred to as “ Cloud Institute of China Academy of communication and communications ”) Jointly published 《 Cold plate liquid cooling 》 and 《 Liquid cooling revolution 》 Two books , The internal and external causes of the development of liquid cooling technology are compared and analyzed , And different liquid cooling technology schemes .
2020 year ,ODCC Open source conference of submerged liquid cooling data center held with Alibaba , Alibaba officially opened the technical specification of submerged liquid cooling data center ;2021 Beginning of the year ,ODCC And UDP( The United Nations Environment Programme - Partner of Danish University of science and technology ) Copenhagen Energy Efficiency Center 、ITU( international telecommunications union ) Jointly issued a report entitled 《Innovative Data-Centre Cooling Technologies inChina-Liquid Cooling Solution》 Liquid cooling Research Report , The application progress of liquid cooling technology in China is summarized .
2021 year ,ODCC Released 《 Submerged liquid cooled server reliability white paper 》, Its research content is the reliability evaluation method of liquid cooling server after long-term operation ;2022 year 4 month , The five series of industrial standards for liquid cooling led by Yunda Institute of China Academy of information and communications have been officially implemented , It provides effective guidance for the development of data center liquid cooling industry .
With the increasing demand for edge computing power , The application scenario of liquid cooling technology will expand from the cloud to the edge . With the increasing demand for terminal computing power , Traditional cloud computing capabilities are gradually sinking to the edge , The power density and heat dissipation requirements of edge computing servers are also increasing synchronously . In order to reduce the cooling energy consumption of the data center , Improve the overall energy efficiency of the data center , Some edge data centers began to introduce liquid cooled edge servers , Improve the energy efficiency of edge data centers through the application of liquid cooling technology .
5、 ... and 、 The research and development of high-density server is accelerated
The growing scale of data and land 、 The intensive development of power resources promotes the development of high-density data center . For a long time in the past , In order to meet the growing data processing needs of users , The data center mainly provides more computing power by increasing space and expanding the size of racks and servers , But this will also lead to the increase of data center operation cost and the waste of data center space .
Increasingly tense land resources in economically developed areas , This makes it difficult to develop the data center construction mode of increasing computing power by expanding the server scale . Building a high-density data center has become an important measure to improve the computing power level of the data center . High density data center can further improve the computing power per square meter of the data center , It can better meet the computing and storage requirements in the big data scenario .
High density server deployment significantly improves the computing power per unit area of the data center , Reduce the operating cost of the data center . The key to building a high-density data center is to deploy high-density servers . Unlike traditional servers , The power supply and fan of high-density server are designed in a shared way , Multiple server nodes in the same chassis can share power and fans .
On the one hand, it reduces the weight and space occupation of the body , The computing power per unit area of the data center is improved ; On the other hand , It can improve the efficiency of power supply and heat dissipation system , Reduce the operating costs of the data center . As the power density of the data center increases , The average rack power of the data center is also increasing . Data released by research institutions show that , At present, the world 5-9kW/ The rack accounts for the most , The power exceeds 10kW/ The rack of the rack accounts for nearly 30%. The improvement of single rack power is closely related to the large-scale deployment and application of high-density servers . at present ,Intel、IBM、 Huawei 、 wave 、 Dawning and other manufacturers have accelerated the product design and market layout of high-density servers .

2020 Average single frame power of global data center in
6、 ... and 、 Enhanced disaster recovery and backup capabilities
The importance of data is constantly highlighted , The demand for disaster recovery backup drives the rapid growth of the backup all-in-one machine market . It is the most important asset of the enterprise stored in the data center . Under the background of accelerating the process of enterprise digital transformation in China , The disaster recovery and backup capability of data center and enterprise computer room has gradually attracted the attention of the industry . Most of the traditional data protection methods can only protect the data of physical devices , It is difficult to provide unified data protection and backup services for cloud computing and other environments . With the growth of distributed computing application scenarios in Data Center , Traditional data protection and backup methods are increasingly difficult to meet the needs . Efficient 、 comprehensive 、 Integrate 、 Unified 、 Flexible data backup and recovery solutions have become the development direction of the next generation data center data protection solutions .
All in one backup is an important solution for the next generation data center data protection scenario . All in one backup machine can use software 、 disk array 、 Server engine or node to back up data , And closely combined with backup software , Make a list 、 Indexes 、 plan , Realize data transmission .
at present , The global backup all-in-one machine market maintains a high growth trend , The market is in great demand . To cater to the development of the industry ,2021 year 9 month ,ODCC Released 《 All in one backup technology specification 》, The architecture of all-in-one backup machine 、 function 、 Non functional 、 The interface requirements have been comprehensively standardized , Effectively promote the production of all-in-one backup machine 、 Testing and selection .

Overall architecture of backup system including all-in-one backup machine
7、 ... and 、 New storage access protocols are evolving , Deep integration of network storage
NVMe-oF The protocol has become an important end-to-end solution for all flash memory , The integration of storage and network is deepening .NVMe-oF It has become a way to reduce the processing overhead of storage network protocol stack 、 An important technology to achieve high concurrency and low latency . It solves the problem of storage performance loss caused by communication protocol , Further released SSD Performance of , Accelerated the development of flash memory .NVMe-oF Enable hosts and storage devices to use through the network NVMe agreement , It effectively improves the performance of host accessing storage through remote network , It has aroused widespread concern in the industry . at present ,NVMe-oF Support FC The Internet 、IB The Internet 、 Lossless network and standard Ethernet . among , Lossless network is one of the focuses in the industry , Great progress has been made in recent years . Its performance is much better than standard Ethernet , The construction cost is lower than FC The Internet and IB The Internet . therefore , stay NVMe-oF In the process of integration and development with the network , Lossless network will become an important development direction .
8、 ... and 、 The demand for ubiquitous computing power interconnection increases , Computing network innovation accelerates
Computing networks connect ubiquitous computing infrastructure , Operators actively carry out landing practice . Finance 、 traffic 、 education 、 Industry and other industries have a growing demand for ubiquitous computing scenarios , Users need more discounts 、 Higher quality computing services that can be accessed anytime, anywhere . Due to the ubiquitous use of computing power and transaction requirements, computing network has become the focus of the development of next-generation computing infrastructure . In recent years , Operators have accelerated the implementation of computing networks . Computing convergence is China Unicom CUBE-Net3.0 One of the key directions , China Unicom has carried out pilot projects of computing network construction in many places across the country , By calculating the network business chain 、 Network slice 、 Resource awareness realizes computing resource scheduling and awareness , So as to form cloud network security integration services .
China Telecom is in the process of computing network construction , Emphasize with “ cloud ” At the core , Focus on Network 、 Integration of computing power and storage resources , Promote the continuous upgrading of Tianyi cloud , Realize the unified management and scheduling of Tianyi cloud and Tianyi edge nodes . Cloud network integration 、 Cloud edge collaboration is the advantage of Telecom cloud . China Mobile focuses on transforming the underlying computing infrastructure , Build on X86 and ARM General purpose computing network based on architecture , Constantly enrich based on GPU and ASIC Intelligent computing power .

Development trend of data center computing network
The innovation of computing network technology mainly focuses on three aspects : framework 、 Scheduling technology and measurement mode . In terms of computing power architecture , The traditional separation mode of computing and network is gradually evolving towards the integration of computing and network , Innovative development of computing infrastructure integration architecture . In terms of computing power scheduling technology , Computing power scheduling has developed from only supporting network scheduling to the scheduling mode that computing power matches the network , Innovations in computing power scheduling technology continue to emerge .
The technological innovation of computing power scheduling mainly includes computing power cross regional scheduling 、 Multi level scheduling 、 Unified arrangement of computing resources 、 Network protocol 、 Visual monitoring and intelligent operation and maintenance . In terms of computational force measurement , The connotation of computing power and network measurement continues to extend , Computational force measurement has changed from single node measurement to integrated collaborative measurement . With the formation of computational power network , The sources of users' computing power have become more diversified . under these circumstances , The computing power available to users cannot be measured by single or local nodes , Instead, we need to use what the computing network platform can provide at present 、 The computing power to meet the business needs of users . This new integrated collaborative measurement mechanism will re verify the available computing power of users .

Blue ocean brain cooling technology
Blue ocean brain liquid cooling cooling server can be built NVIDIA 4 × A100 / 3090 / P6000 / RTX6000, The liquid cooling plate is fixed on the main heating device of the server , Rely on the liquid flowing through the cold plate to take away the heat to achieve the purpose of heat dissipation . Cold plate liquid cooling solves the heat dissipation of devices with large heat generation in the server , Other radiator parts have to rely on air cooling . Therefore, the server using cold plate liquid cooling is also called gas-liquid dual channel server . The liquid of the cold plate does not contact the cooled device , Heat conduction plate is used in the middle for heat transfer , High safety .

Blue ocean brain has the ability of completely independent research and development and flexible customization , It is the first one engaged in liquid cooling in China GPU One of the manufacturers of servers , And in the unique reinforcement 、 Deep accumulation in product technology fields such as confidentiality and security . Blue ocean brain end-to-end solution , Empower AI、 Cloud computing 、 big data 、5G、 Blockchain 、 Innovation and application of emerging technologies such as yuancosmos , Actively contribute to China's new infrastructure construction 、 Xinchuang 、 Count from east to west 、 The digitization and intelligent transformation and upgrading of social economy provide a solid guarantee of computing power . At the same time, we attach importance to global cooperation and win-win results , Intel 、 Ying Wei Da 、 Seagate 、 Huawei 、 Feiteng and other domestic and foreign industrial chain partners maintain close strategic cooperation , Accelerate the adaptation of product solutions and the expansion of application scenarios , Better serve users in various industries .

END
attach :《 Data center white paper 2022》 Part content :






Add a little assistant wechat , Enter the industry group , obtain 《 Data center white paper 2022》 The full report .
边栏推荐
猜你喜欢
随机推荐
Neo4j 知识图谱的图数据科学-如何助力数据科学家提升数据洞察力线上研讨会于6月8号举行
以不太严谨但是有逻辑的数学理论---剖析VIO之预积分
求1~100之间不能被3整除的数之和
液冷数据中心如何构建,蓝海大脑液冷技术保驾护航
Chain stack
Interpretation of the paper: using attention mechanism to improve the identification of N6 methyladenine sites in DNA
Introduction and use of Ninja
The green data center "counting from the east to the west" was fully launched
Notes | (station B) Adult Liu: pytorch deep learning practice (code detailed notes, suitable for zero Foundation)
All kinds of ice! Use paddegan of the propeller to realize makeup migration
Find the sum of numbers between 1 and 100 that cannot be divided by 3
MySQL transaction
Vio --- boundary adjustment solution process
静态链表
Hard disk partition of obsessive-compulsive disorder
Interpretation of yolov3 key code
Ninja file syntax learning
numpy总结
Ffmpeg audio coding
VIO---Boundle Adjustment求解过程









