With the century-old topic of energy saving and emission reduction, green data centers are gradually heating up. The sustainable development of the environment is becoming an increasingly priority issue for the IT industry. The development of information technology cannot be left without "green", but also subject to the constraints of "green" and conducive to the realization of "green". At the same time, the concept of green data center Proposal and improvement are inseparable from the development and support of information technology. A data center greening and even the greening project of the entire IT architecture is in full swing.
Energy consumption and space utilization are the biggest cost of the data center and the main factor restricting the expansion of the data center. The new generation of green data centers can help companies save costs, improve service efficiency, quickly respond to business needs, better match business goals, and undertake energy saving , Environmental protection and other social responsibilities.
The meaning of "green" mainly refers to two aspects. One is the infrastructure construction of the data center computer room, limiting the use of harmful building materials, and the design of the computer room must be thermal insulation, thermal insulation, energy saving, radiation prevention and mute. On the other hand, IT equipment itself should limit the use of harmful materials, reduce power consumption, reduce volume, save energy, mute, etc. We regard the comprehensive use and implementation of measures such as restricting the use of harmful substances, reducing energy consumption, saving space, improving equipment efficiency, and mute as the basic connotation of green data center construction. In other words, the green data center refers to the standard design of lighting, computer room equipment, power and IT systems that follow the maximum energy savings and the lowest environmental impact, while at the same time improving the utilization rate of each equipment and having a good return on investment.
In the construction of green data centers, it is easier to do without the use of hazardous substances. This should be paid attention to and the selection of non-toxic and non-hazardous substances and materials in architectural design, computer and server architecture design. Silence can also be achieved through shielding and sound insulation. The problem lies in the increasing demand for server capacity, the increase in the number of servers, the expansion of occupied space, and the power consumption of data centers is soaring. Therefore, reducing energy consumption has become a key issue in the construction of green data centers.
According to IDC global research, the current installed capacity of the entire server is about 35 million units, and the annual electricity consumption of the server is about 50% of the annual server purchase cost, which is about 29 billion US dollars; the cost of energy consumption has become 30% to 50% of the operating costs consumed by the IT industry. The total power consumption of China's IT products is estimated to be 30 billion to 50 billion kWh, which is almost equivalent to the total annual power generation of the Three Gorges Power Station! And the power consumption of data centers is still increasing at a rate of 15% to 20% per year. The amazing data tells us that IT energy saving is urgent!
In order to complete the data processing and analysis tasks of enterprises, large data centers are equipped with dozens, hundreds, or even thousands of servers, and a large number of disk arrays for servers to store data. In order to work together, these devices They also need to be cascaded through the network, and they are managed with collaborative software and equipment, which results in numerous equipment, low efficiency, and a large footprint. The power supply, mute, air conditioning, and heat dissipation of the computer room are all problems. One by one is displayed. Faced with many problems to be solved, mainly to improve the power utilization rate of IT equipment (servers, storage, network equipment, etc.) and minimize the heat loss of power. In addition, while increasing the computing density, through new heat dissipation technology and data The overall design of the center solves the heat dissipation problem.
The support system and operating environment of the data center involve a wide range of professional fields, covering power, electronics, computers, machinery, chemical industry, thermal energy, etc. The construction of a green data center is definitely a systematic project, and we will try our best to elaborate and solve these problems. All aspects of the problem.
Design overall goals and overall plans, and propose overall solutions
First of all, we must understand and analyze the specific situation of data center applications, propose an overall solution to these problems, and plan well according to the company's development. It is best to plan the computer room comprehensively, and then implement it step by step according to the needs of business development. Only in this way can the overall goals of improving the efficiency of the data center, reducing power consumption, and reducing floor space be achieved.
Today, when the active construction of the data center becomes more and more clearly our IT construction task, we need to calm down and analyze the various needs of the data center construction, and paint a blueprint for the development of the data center, so as to provide our future Work sets direction. To maintain a clear development strategy and direction, and develop in accordance with the construction blueprint, new, highly productive, forward-looking, and highly flexible new data centers will escort the development of enterprises under the increasingly fierce market competition environment.
Therefore, we need to model the construction of the data center. What needs to be considered here is the construction of a data center for complex tasks. In order to ensure the simplicity of the data center, we need to abstract the tasks and perform generalized processing for this purpose. Carry out detailed analysis and comparison of different data types and different data management functions under different storage architectures, and fully consider the requirements of business processes to achieve the construction purpose of "easy management, safe and reliable, flexible expansion, and economical benefits". A modern data center will be an "automated green data center."
Greening of data center buildings
In terms of energy saving of the building itself: In order to reduce the heat exchange between the data center computer room and the outside world, to save energy consumption, and also to improve safety, all the building walls of the computer room area are not provided with external windows to reduce solar radiation heat into the computer room. Reduce the heat loss of the room; the external wall structure of the building adopts high-level thermal insulation design to reduce the cooling loss of the data center computer room; the internal wall and floor of the building adopt thermal insulation design.
Reasonable layout of data center computer room layout
The construction of various systems in the computer room depends on the building environment and is also constrained by factors such as the building structure and plane layout of the computer room, such as building height, shape, and area. The layout of the main building of the data center computer room focuses on functional design. The architectural design is designed to be simple and practical. It consists of the basic building blocks of the data center computer room, other auxiliary rooms, fire stairs, passages, freight elevators and other areas.
The data center room height of the data center should fully meet the laying of the ground air-conditioning air supply and ceiling ceiling return air and related pipes (air-conditioning water pipes, fresh air pipes, fire exhaust pipes, and gas fire-extinguishing pipes) in the data center.
High-density area design and high-performance computing technology
In the data center, a high-capacity cabinet area can be set separately in the computer room area to meet the future development needs of high-density equipment. Improve unit space computing capacity, integrate computing nodes, use one machine for multiple purposes, and reduce storage space requirements. In addition to the system layout, related supporting systems (such as power distribution switches, cables, etc.) are designed accordingly to ensure the integrity and matching of the system design. The cooling problem of high-density equipment is a key consideration. The high cooling demand is a great challenge to the design and layout of precision air conditioning systems. The height under the floor, the air conditioning air supply distance, the air conditioning cooling capacity design, the return air flow design, the cabinet design, etc. are all considered, in order to ensure the overall The functional plane is fully optimized under the premise of sufficient cooling capacity.
High-density and high-performance computing technologies are inevitable trends in the development of IT equipment.
The design of the high-density area solves the problem of power supply and heat dissipation of high-density equipment, and effectively solves the dilemma of local hot spots in the data center.
Efficient air flow structure design
Efficient air flow design will optimize the cooling effect of the entire computer room and minimize the energy consumption of the air conditioning system. Electronic equipment in large and medium-sized data centers is densely arranged, and the total cooling load is relatively large, about 300 to 600W per square meter, and some are even higher. Among them, the equipment cooling load accounts for more than 80%. In view of the characteristics of large residual heat and concentrated heat sources in the computer room, it is necessary to have a reasonable distribution and distribution of airflow organization to effectively eliminate the residual heat in the computer room to ensure that the electronic equipment has a temperature, humidity, cleanliness, air supply speed The need for comfort for personnel. The airflow organization form of the equipment room includes bottom-up and top-back, side-up or bottom-back, and diffused air supply. The airflow organization form should be carefully considered, and the efficient use of cooling capacity should be considered according to the equipment cooling method and installation method Keep the heat sink within the range of cold air jets.
According to the center, a dedicated air supply channel shall be constructed. The equipment in the data center has a large energy consumption, a large amount of heat, and the best effect of the downward air supply mode. Therefore, it is necessary to install a floor in the equipment room and establish a dedicated downward air supply channel.
Highly reliable and highly available intelligent power supply system design
For the data center, the "green" degree of the power supply system is also crucial.
In terms of power distribution system energy saving, high-voltage generator sets and high-voltage direct-to-floor designs can be used. The highest availability UPS power supply system that does not need to convert the load to the city power at all can be established. The UPS output is as close as possible to the final load, reducing line power loss, scientifically managing and suppressing harmonics, reducing transformer and generator losses, using high efficiency and harmonics Low wave UPS equipment, such as APC's InfraStruXure solution uses modular intelligent UPS, its modular design ensures N + 1 redundancy and expansion capabilities, and excellent quality UPS equipment has voltage stabilization and surge protection functions.
Using high-efficiency new gas light source as outdoor lighting, indoor energy-saving lamps, using automatic measurement technology, precision intelligent power distribution design, rack-level power distribution monitoring, monitoring each branch voltage, current and capacity, and through the network Realize remote control to achieve on-demand distribution and supply of power resources. DC power supply is mostly used, because the efficiency of the DC system is 20% higher than that of the AC system. Huawei's full range of telecommunications equipment, storage equipment, security equipment, server equipment, and network equipment all support DC power supply.
High-performance cooling system
Cooling technology has been developing over the past 40 years or so, but it has failed to keep up with the technological innovation of IT equipment. When traditional data center cooling technology is facing the challenge of cooling modern high-density data centers, it seems a bit overwhelmed. Hot air may mix with cold air, and the system cannot identify hot spots, causing hot spots in certain areas. The power density of a modern data center with IT equipment is very likely to reach 15 kilowatts per rack, and non-traditional cooling should be considered. A dynamic intelligent cooling concept is worth recommending. This concept refers to creating an environment in which the cooling system equipment gets continuous feedback from the environment and dynamically adjusts itself to adapt to changes in environmental parameters. These parameters are monitored. This type of system is best suited for today's data centers and helps reduce power consumption, which is undoubtedly a boon for data centers. Another concept is natural cooling. If your data center is located in a geographically appropriate location and the weather permits, the best way is to use the natural cooling available in the area. This can be accomplished through the use of economizers that supplement the main cooling of the data center through the existing natural cooling outdoors. These saving devices can run air or liquid, because it saves electricity, it is very worthwhile. In some areas, as long as the average daily temperature is less than 10 degrees per year, the data center air conditioning host does not need to be activated, and uses the outside natural cold air to cool the entire computer room environment. If your data center is being designed from scratch, one of the best options is to provide elevated cooling instead of the traditional under-floor pressure system. This arrangement may save up to 30% in energy costs.
In terms of energy saving in air conditioning systems, the data center computer room can adopt a centralized cold water supply cooling design, select high energy efficiency ratio cold water host equipment, adopt a pump variable flow system design, adopt an energy-saving air conditioning system design, and use a cooling water pump redundancy design to improve system reliability. Efficient cold and heat pretreatment fresh air system, the data center computer room adopts centralized humidification, cancels the precision air conditioning humidification function, and reduces energy consumption. Using the water cooling method, the same volume of water can carry 3500 times the heat of the same volume of air. Another water-cooling method uses a complex external cooling system to cool the electronic components through a closed-loop water cooling system in the rack, and the water supply and drainage pipes are covered throughout the data center. This compact system can reduce the air flow path, thereby reducing the fan's energy consumption. This compact heat dissipation method minimizes or even completely eliminates the mixing of hot and cold air.
APC is based on rack-level InRow cooling technology. Compared with traditional room-level cooling, it can effectively ensure that core switches and blade servers are efficiently and accurately cooled for heat load transfer, which reduces energy consumption and reaches a higher level. Cooling effect. Although the emergence of blade server architecture has greatly improved the capacity of the server, due to the intensive integration, it has brought about the problems of ventilation and heat dissipation. The imported cooling method can successfully solve the heat dissipation problem of the blade architecture.
In addition, the overall monitoring of precision air conditioners should be strengthened, and the operation can be adjusted at any time as the environment and humidity change.
IT microenvironment
Compared with the computer room environment, the IT microenvironment is the real operating environment of IT equipment. The IT microenvironment refers to the environment on which the IT equipment in each rack operates and includes key elements such as power, air conditioning, and monitoring. Innovatively designed cabinets fully consider the power distribution, air distribution, load-bearing and wiring requirements of IT equipment to meet the reliable operation of IT systems.
The IT microenvironment is the only way to create high availability. To improve the usability of the overall IT system, we must start with improving the usability of the IT microenvironment. Among the many key elements involved in the IT microenvironment, the power supply protection subsystem has been recognized and adopted by most users. However, the temperature regulation, heat management, cable management and other links in the rack have not received enough attention, and they do not really provide a highly available operating environment for the IT equipment in each rack.
Generally, due to the rapid increase in the number of server devices in the data center, in order to facilitate management and stacking, more and more servers are racked, and the number of blade servers is also increasing. Although this method saves space, the internal environment of the rack with increasing power density is like a time bomb. It is important to know that in addition to power factors, the stability of the system is related to heat control. According to tests by the authoritative power protection research center, every time the temperature inside the cabinet rises by 10 degrees, the life of IT equipment will be reduced by 50%. The problem of cooling cannot be solved well, and the damage to IT equipment is quite huge.
The traditional heat control method is to reduce the temperature of the entire computer room, and it is expected to achieve the function of "cooling" the server. But it turns out that this method can only control the temperature of the entire room, and can not guarantee the temperature of the server in the rack. A more scientific solution is to use the rack as a unit to control the temperature of the IT microenvironment. By reasonably setting the cold and hot aisles, the cold wind is directed to the air inlet of each IT device, and the hot wind is effectively sent back to the air conditioner to ensure that the temperature in the rack is controlled to the temperature suitable for server operation.
In addition to thermal management, cable management in the computer room is also one of the factors that must be considered to improve availability. We can't just focus on the clear requirements for the wiring of the computer room as a whole and the local area network. In this particular IT microenvironment in and between cabinets, the messy cables greatly reduce the efficiency of the management staff of the computer room. Once the network fails, fault finding and troubleshooting will become a nightmare. The power distribution and wiring of the computer room are often not paid enough attention in the initial stage of the establishment of the computer room. When it becomes a problem, they will be found to have seriously affected the availability of the entire computer room.
The cable management of each rack, including power distribution, wiring and other details, may become the fuse of future IT system failures. The easiest way to avoid all problems at once is to adopt a holistic solution that integrates rack, cooling, power, management and maintenance into a whole.
Because the overall solution eliminates the need for users to pay attention to the details of IT systems, the effect of improving usability is obvious and less difficult. This kind of overall solution to design IT microenvironment construction from the standpoint of overall usability has become the general trend.
Select IT equipment with high availability and low power consumption
The selection of high-availability and low-power IT equipment should be considered from the CPU, memory, and all levels of the entire equipment.
In addition to the low-power design of chips advocated by processor manufacturers, server manufacturers have launched various "fresh" green products to help large enterprises with data centers improve energy utilization and improve data center design solutions. Thereby reducing the energy cost of the data center and even lower the total cost of ownership. This has indeed become an objective demand, the needs of the times.
Using multi-core X86 chip technology can improve the processing capacity of the data center. Conventional data centers require hundreds or thousands of servers. The main problem is that the server capacity cannot meet user needs. Servers with multi-core X86 chip architecture have greatly improved their capabilities, while the complexity of the system has been greatly simplified as the number of servers has decreased. In the storage system, under the premise of ensuring performance, the hard disk with lighter weight, higher reliability, and lower power consumption and running cost should be selected.
The biggest loss of the server does not come from the inside of the chip, but from the conversion of AC to DC. The data shows that the efficiency of AC to DC conversion is generally 60 to 65%. The energy consumption outside is obviously directly turned into heat. For data centers, this "excess" heat brings another problem, that is, heat dissipation. Large and medium-sized data centers basically spend a quarter of the cost on cooling. In this way, the AC energy of the server is converted into the heat energy produced by the DC power, and it needs to spend energy to dissipate them. Although the blade server power consumption is much lower, so that the computing density has increased significantly, but also to solve the problem of heat dissipation. Blades will be replaced by more flexible server types in the future. The computing architecture of such servers will treat memory, processors, and I / O devices together as shared resource pools.
Transform the overall architecture for system optimization
Simplify the complexity of the physical configuration of the data center (server, network, storage, etc.) and reduce the footprint. In order to ensure the completion of their data processing tasks during peak hours, the data center has deployed dozens, hundreds, or even thousands of servers. For these servers to work together, they are connected through the network and shared storage. Only when equipment and software manage them, can the system coordinate work and complete data processing tasks. When I visited these data centers, I felt that the system connection is very complicated, and the wiring is very chaotic and difficult. The overall architecture of the data center needs to be changed. In fact, the server architecture currently in use has been used for decades, and problems have surfaced. The advent of blade machines has taken the server architecture design a step forward. But it also brings about the problem of ventilation and heat dissipation. Although many network devices and storage devices in the data center are connected through the network, they are actually still an independent group. The integration and optimization of data center equipment, intensive and integrated design have been mentioned in the agenda. To improve the efficiency of the server and the efficiency of the entire data center, to save energy, reduce consumption and reduce the overall cost of ownership, this requires that the architecture design of the data center and server must be changed, and users are expecting the arrival of the new generation of servers.
After system optimization, Egenera integrated these needs into a resource pool, which is connected to each user for use like tap water. After such integration and optimization, the connection cable at 24 servers is reduced to about one-eighth of the original structure. The server footprint is reduced to about a quarter compared to blade servers of other manufacturers. The area occupied by the intensified combination of these devices has also decreased.
Virtualization technology improves equipment utilization efficiency and reduces power supply cooling cost
The basic concept of virtualization is simple. It encapsulates computing resources and runs a shared physical infrastructure, so that each device behaves as if it is operating in its own independent physical environment. Through virtualization and consolidation, using storage and computing resources as an aggregate storage pool can achieve this goal, and users can access networks, systems, and applications based on demand. In addition, through measurement, we can find that usually a server only uses 5% to 15% of the total capacity to provide application services. Using virtualization technology to intensively integrate multiple blade servers into a virtualized server platform, and using virtualization technology to integrate multiple server virtualization platforms greatly improves the efficiency of server utilization. Using virtualization technology, the integration of underutilized servers is seamless for end users. The use of virtualization and intensive technologies improves the efficiency of server utilization and greatly reduces energy consumption. Server virtualization can greatly improve the operating efficiency of IT loads and save energy consumption. Virtualization can help the effective logical integration of servers, reasonably control the scale of hardware, improve the utilization of server resources, and improve the operating efficiency of server CPUs, and virtualization can also provide good support in backup, storage, and downtime. By providing faster backup and recovery speeds, a complex test environment can be built, which saves machine room and floor space, and reduces the energy consumption of cabinets, network cables, and air conditioners. Virtualization technology is an important symbol for promoting green applications in data centers. Virtualization technology that uses fewer hardware devices and power supplies to achieve greater processing power will be widely used in the future.
The virtualization mentioned two or three years ago refers to the virtualization of a product line, and now the virtualization refers to the virtualization of the entire data center, including servers, storage, network and other products will be virtualized .
The application of virtual technology can alleviate the contradiction of low performance utilization of servers, but when many servers form a data processing platform, how to play the role of virtual technology, related technical issues still need to be studied and resolved. Comprehensive use of blades, virtualization, water cooling, power execution and other technologies can reduce the power consumption of data centers by 40-50%.
Data center needs comprehensive management design
The new-generation data center management system not only requires the safe operation and management of each device, but also requires the computer room management system to provide data related to cost operation to meet the relevant energy-saving specifications, automatic measurement and automatic monitoring, and customers only pay for their own resources ". These data centers will be based on the comprehensive management of service-oriented architecture (SOA), providing the computing power and intelligently-supplied IT capabilities required by users on demand, just like SaaS operating management and providing services. The PUE (power usage efficiency) indicator currently in use in the data center industry is an indicator that shows the actual power consumption demand of computing / IT components (such as servers). To calculate the PUE of a facility, simply use IT equipment to consume power The amount can be divided by the power consumption of the entire facility. The number of results can be between 1 and infinity. The closer to 1, the better your PUE. PUE management should be considered for green data centers. PUE is the best standard to solve the energy efficiency of data centers.
Modern data centers require the system to provide more practicality, flexibility and scalability than before. Today, you need much more than server management, but a system that guarantees the normal, safe, and stable operation of the entire IDC. Centralized data center management provides you with access and control of all IT assets in any location and in any state, and allows you to complete access and control of the data center without visiting the site, which greatly facilitates your overall management. . The data center's remote centralized management and control system provides IT professionals with access to equipment in the data center wherever there is an IP network, as if the equipment were nearby. This approach not only speeds up response time and shortens downtime, but also improves the availability of IT equipment, and allows managers to manage freely whenever and wherever they are. The data center remote centralized control solution allows multiple operators to perform remote centralized and unified operation and control of servers, minicomputers, blade servers, virtual hosts, and network devices in the data center at the same time, and can record the operator's complete control of the target device throughout the process. The content of the operation screen is used to track and standardize the operation behavior of the operator. Not only can the common and unintentional misoperations of the operators be discovered in time, and the post-mortem analysis and correction can be performed; the malicious operations of the operators can be traced back to provide first-hand evidence of malicious operations.
The construction of green data centers has grown into "green shoots", and some key technologies that support and drive the growth of green data centers are constantly being developed and improved. Green data centers are thriving and blooming and bearing fruit. Haode Technology focuses on promoting the industry norms and national standards of China's new generation of green data centers, hoping that the new industry norms and national standards can guide and promote the upgrading of China's green data center industry. We will launch systematic solutions based on the latest international standards, the most advanced green data center design concepts, optimal construction experience and operational management specifications. Become a domestic first-class comprehensive IT service group centered on data center infrastructure and oriented to "disaster recovery + production" integrated outsourcing.
