What is a Data Center?

A data center is a global network of specific devices that are used to transfer, accelerate, display, compute, and store data information on the Internet network infrastructure.

A data center is a network of specific devices that work together globally to transfer, accelerate, display, compute, and store data information on the Internet infrastructure. Most electronic components in data centers are driven by low DC power. ^[1]

With the rapid development of the data center industry in the world and the rapid growth of society and economy, the development and construction of the data center will be in a high-speed period.

Common communication failures in data center networks are mainly concentrated in:

(1) Analysis of failure phenomena:

In general, due to the complexity of the components, the failures also show different manifestations. So if you want to analyze the fault, you must first understand the phenomenon of the fault. For example, if there is a problem in the payment system that cannot be paid by the payment system, and the webpage is difficult to open, then you must check the relevant points of failure one by one. And other devices. Therefore, it is necessary to collect and sort out several common faults of the data center network, and perform retrieval and search according to the phenomenon.

(2) Test and confirm the fault scope, and locate the fault point.

All application services are carried out on the basis of the normal operation of these physical hardware, and some hardware problems will cause failures. According to the performance of the fault, it is necessary to perform screening and inspection for each part, for example, testing the server and inspecting network equipment. According to the performance of the problem, one by one is eliminated, and the location of the fault point is finally determined.

(3) If all the above hardware failures have been eliminated, then it is the failure of the computer system. This failure requires the establishment of a failure model for diagnosis and definition according to the PMC model. Through the layered test method, find the problem unit, that is, normal unit test normal unit, normal unit test failure unit, fault unit test failure unit, fault unit test normal unit and so on. The detection results of the latter three types are all faults, so you can establish a limited number of units through hierarchical measurement, and use the matrix and the fireflies algorithm to focus on the FAFD algorithm to diagnose other units, and finally determine which system or systems are faulty Units. Of course, you can also use mirroring, traffic statistics, packet capture, and other methods to determine the scope of the device where the fault is located, and then narrow the scope to focus on one or more devices.

(4) Collect important data information.

When troubleshooting, collect information such as equipment logs, diagnostics, and operation records to summarize these data. If conditions permit, set up a fault database. For common problems, you can "handle as soon as they occur." There are no failures, you can continue to collect into the database. In short, the necessary information collection will help to better find the cause of the failure in the future and ensure the healthy and stable operation of the data center network. ^[2]

In the information age, data centers have also brought convenience and economic benefits to more companies. For example, Tencent s QQ, almost every mobile phone user has

The development of data center cooling technology has the following trends.

First, the end cooling equipment is close to the server

Data center computer rooms used room-level air conditioners and cooling air under the floor. This method has low construction cost and high utilization rate of the equipment room, and is used to solve the heat generation of a single cabinet of 3 to 5 kW. However, with the large-scale application of rack-type and blade-type servers in computer rooms, the number of devices in a single cabinet, power density, and heating density have increased significantly. Traditional computer room air conditioners have been unable to solve the heat dissipation problem of IT equipment, and row air conditioners and backplane air conditioners have emerged as the times require. The end of this new type of air conditioner is closer to the heat source, which can solve the problems of local hot spots and high heating density. Through the close-range cooling transmission, the power consumption of the fan is reduced and energy saving is achieved. Regardless of room-level air conditioners, row-level air conditioners, and backplane air conditioners, the air is cooled first, and then the cold air is used to exchange heat with the CPU of the server to cool down. Due to the low heat exchange efficiency and heat flux density of air, air-cooled servers have problems such as high cooling energy consumption, high noise, and low equipment density. In order to solve the heat dissipation problem of ultra-high power density IT equipment, data centers have begun to use liquid cooling technology, using working fluid as an intermediate heat transmission medium, to transfer heat from the heating area to a distance and then cool it. Liquid cooling technology has significantly higher cooling efficiency than air cooling, which can effectively solve the heat dissipation problem of high-density servers, reduce the energy consumption of the cooling system and reduce noise.

Second, attach importance to energy saving of cooling system

As competition in the data center intensifies and the pressure on operating costs increases, research into energy conservation of cooling systems is imminent. With the advent of high-temperature servers, more and more data centers are gradually trying to develop and use free-cooling. The use of natural cold sources mainly has the following two directions.

1. Fresh air direct cooling

A typical case is Facebook's data center in Princeville, Oregon. After the outdoor fresh air is filtered, it is humidified and cooled, and then sent to the air inlet of the rack through the fan wall. The outdoor fresh air is heated by the server and discharged to the outdoor. This method has high requirements for outdoor air quality. At present, some data centers in Ningxia Zhongwei have tried to apply this technology.

2.Using deep lake and river water for cooling

A typical case is the Dongjianghu Data Center in Zixing, Hunan Province. Dongjiang Lake is the largest man-made lake in central and southern China, with abundant cold water resources. Dongjiang Lake covers an area of 160 square kilometers and has a water storage capacity of 8.12 × 109 cubic meters. Its downstream Xiaodongjiang water temperature is below 10 all year round, the water flow is stable and the water quality reaches the national first-class standard. 90% of the year does not require mechanical cooling, and river water direct cooling is used. The average annual PUE is expected to be 1.15. ^[3]

The main purpose of a data center is to run applications to process the data of business and operational organizations. Such systems belong to and are developed in-house or bought from enterprise software vendors. Like general applications there are ERP and CRM systems. A data center may focus only on the operating architecture or provide other services. Often these applications consist of multiple hosts, each host running a single component. Usually such components are databases, file servers, application servers, middleware, and various other things. Data centers are also often used for backups at offsite sites. The company may order the services provided by the data center. This is often used in conjunction with backup tapes. Backups can place server-local stuff on tape, however, tape storage locations are also vulnerable to fire and flood security threats. Larger companies may send their backups to off-site locations. This can be done by the data center through re-investment. Encrypted backups can be sent to another data center over the Internet for safekeeping. For disaster recovery, various large hardware vendors have developed mobile device solutions that can be installed and operational in a short time. Vendors develop systems like Cisco Systems, Sun Microsystems, IBM and HP that can be used for this purpose.