How to build a cloud computing data center?

Cloud computing is about to become a public resource for the information society, and data centers are the infrastructure that supports cloud computing services. How to build and operate data centers for cloud computing is becoming the most important task for China's communications and IT industry in the past three years. One of the fields.

What is a cloud data center

What is a cloud computing data center? What is the difference between his traditional IDC (Internet data center) and EDC (enterprise data center)? The difference between a traditional data center and a cloud computing data center lies in the different business models. Most traditional IDCs support telecom operators' data services and have clear cross-network and regional restrictions. The information system architecture supported by traditional EDC is also different from cloud computing. For example, EDC supports the application-specific information system based on commercial software. Therefore, its scale, level, and variables are relatively fixed. The data center needed for cloud computing comes from the Internet, but it evolves toward the integrated platform. Therefore, it is different from the situation where the traditional data center infrastructure and information systems are separated from each other. The data center of cloud computing is from infrastructure to computing and Applications are continuous and integrated, and they are related and adaptable.

How to build a cloud data center

How should a cloud data center be built? At the moment, there are two common ideas in the industry.

Point of view: "High-investment, high-performance supercomputing is cloud computing."

Point two: "Cloud computing is to put together a bunch of rotten machines to provide highly reliable services."

These two views correspond to two different modes of construction.

The point of view corresponds to the traditional construction model. This model focuses on building a high-cost and high-reliability infrastructure and then implementing a simple virtualized private cloud with physical boundaries. For example, banking data centers are simply stacked at cost, and efforts are made to improve the reliability of the infrastructure. However, due to the separation of data center hardware and software systems, the overall efficiency is reduced under high investment.

The second model corresponding to viewpoint two features the blind pursuit of reducing the initial investment cost, the data center infrastructure as a simple resource construction and operation, and the establishment and operation of a low-cost, low-efficiency data center infrastructure, resulting in initial costs. Although it is low, the operating costs and operating risks are relatively high. Although this adapts to the current rapid development of the industry, it is generally not suitable for the long-term development and stable service of cloud computing.

The author believes that both cloud data center construction models have one-sidedness and limitations. Like the early years of the Internet, although it can be carried on the PSTN (Public Telephone Network), it is not the best model. The infrastructure built for cloud computing can run on IDC, but it cannot effectively optimize and develop cloud services. For public facilities that serve the community, cloud computing must re-plan and establish infrastructure such as data centers.

The cloud computing infrastructure should be fully integrated with the computing platform. Just like the classic book "The Datacenter as a Computer" in the cloud computing industry, the cloud computing data center is a top-down, full integration and optimization process. Including system software, chips, storage, networking, power, power distribution and cooling, a new generation of cloud computing data center fully integrates software, servers, networks, data center power supply and cooling, energy and other The system and complete system optimize the overall efficiency from energy to computing. For example, Google did not narrowly optimize data center PUE (energy use efficiency), but based on software, it incorporated a custom infrastructure and established a global-scale cloud computing infrastructure.

In implementing the specific construction of the cloud data center, the author believes there are the following points of attention.

First of all, in terms of business models, infrastructure needs to be fully adapted to the objective laws of cloud computing business development, established in accordance with the unit size of the most optimized efficiency, and efficiently build according to business development needs, abandon the traditional model, it is best not to build large-scale one-time The data center is a monolithic multi-story building, and will adopt a flexible and efficient modular approach to completely divide the distribution of data centers, cooling, power supply, and even buildings and mains electricity, with a single cluster of cloud computing services as the smallest unit. Implementation.

In terms of data center availability, it is advisable to optimize the structure based on the characteristics of the cloud computing IAAS system, and to achieve adjustments that can be adjusted according to business needs. The availability of second-generation or even lower (traditional IDC cannot be achieved) modular infrastructure infrastructure availability and operation can be improved. cost. In the range similar to the T2-T4 level, the quality of the SLA (Service Quality Commitment) that can guarantee the data center infrastructure of the five nine levels of the year is provided. At the scale of the smallest unit (granularity breakdown), it will be appropriately matched with the cloud computing system, allowing the deployment of fewer physical server clusters at a time and also providing an efficient system.

In the cloud computing business service model, we must use infrastructure as a platform to implement data center accurate energy billing, separate assets from electricity or energy, and achieve optimal cloud computing resource operating efficiency and the best economic model.

In terms of data center facility delivery efficiency, cloud computing services have higher requirements. With the rapid development of the industry, system delivery will soon be required in months or even weeks. This is in contrast to the traditional annual delivery cycle. The big difference is that the custom-built, modularized and highly coupled business infrastructure construction model will greatly increase the efficiency of initial production, shorten the first implementation cycle to less than 12 months, and control the production efficiency (expansion/addition) to three Within months.

When constructing a cloud computing data center, it is best to also build data centers in different regions in the same region based on the requirements of leading cloud computing technologies, and ensure reliable T-level high-speed interconnection to achieve the reliability and availability of cloud computing architecture. The best match is to effectively reduce the impact of regional disaster factors on the destruction of data center infrastructure and the impact on cloud services. By aggregating the resources of regional data center clusters, a stable platform for cloud computing systems will be established to achieve data center intensive and large-scale management, reducing the complexity and uncertainty of operation and maintenance.

In terms of economics, cloud computing data center cost control and cost-effectiveness are critical. The cloud computing data center will focus on the reduction of total cost (TCO) and focus on optimizing operating costs (Opex) while measuring cost/efficiency/availability The balance.

In the development of the industry and business cooperation, cloud data center service providers must identify the positioning and service models, take the data center infrastructure as the core business, and actively cooperate with the telecom operators, and use their respective modes of resource sharing and business win-win. Its resources, cost and service advantages, develop and serve the public cloud business in China.

In short, the establishment of a new generation of cloud computing infrastructure should aim at the high efficiency, low operating cost, flexible business adaptability and service availability of cloud computing data centers, and be implemented and implemented in phases.

Cloud Data Center Location

Finally, we discuss the location of the following cloud data centers. The author's point of view is that the construction of cloud data centers should go west all the way. Because western China has more value points for the establishment of large-scale cloud computing data centers, it is relatively clear and significant that it is energy, climate and geographical location.

Taking the Ningxia Autonomous Region as an example, Ningxia's geographic region has a strategic and strategic position covering the whole country. Ningxia is located in the geographical center of China, covering the same distance between the eastern and western regions. Ningxia is located 1,000, 1,600, 1,700, and 800 kilometers away from the north of the city. It is within 15 ms of network communication. For cloud services, the space is not the distance and the delay is the distance. After the large-capacity network is established, the resources of Ningxia can be used by the West and East, and the transmission power can be changed to be the transmission bit. The industrial structure has obvious advantages in matching. Secondly, Ningxia is located in the back of the inland region, and is located in the depth of the national security strategy. At the same time, its relatively rich population and abundant environmental capacity are the ideal environment for the establishment of strategic data reserves and processing.

The climate in western Ningxia is characterized by a typical frigid continental climate. The annual average temperature is octaves, and the climate is dry and cold. The average temperature of minus ten degrees is more than 180 days. The annual maximum and minimum temperature difference is small, and evaporation is more than ten times of precipitation. Under the condition of a reasonable climate site, nature is the best energy efficiency tool in the data center. By using the climate in this area, evaporative cooling mode is adopted, and the cleanliness of the air is matched with high cleanliness of the air, less corrosive substances, and less dust particles (results of sand control in western Ningxia ), Data center energy efficiency can be effectively improved to achieve synchronization with the best level in the world.

Ningxia has abundant electric power resources, low energy costs, and is an important oil and gas energy channel and power transmission base in China. Ningxia currently has the lowest on-grid electricity price in the country, and the scaled photovoltaic and wind power renewable energy sources in the western part of Ningxia are among the highest in the nation in installed capacity and grid-connected rates. The ratio of green power to the power grid is high, and the power supply is stable and reliable (Ningxia is the transmission export of Qinghai and Gansu). These resources and conditions are to provide large-scale cloud computing data centers that provide economic guarantees for long-term, reliable, and reliable energy resources, and to integrate with international green energy and cloud computing.

At present, cloud service providers have already chosen to establish large-scale cloud data centers in western provinces such as Ningxia. It is believed that in the near future, China's central and western regions will become a gathering place for cloud data centers.