When it comes to how the “East Data West Calculation” strategy is a grand chess game,
we must mention data centers.
A data center, abbreviated as IDC, stands for Internet Data Center.
The reason it is not directly referred to as “DC” is mainly to avoid confusion with Direct Current. Moreover, current data centers are generally connected to the internet and primarily serve internet businesses, making “IDC” a more accurate term.
Functionally, a data center is a super-sized server room filled with numerous servers dedicated to centralized management of data (storage, computation, exchange).
According to industry statistics, in 2020, the global data traffic processed by data centers reached 15.3ZB (1ZB ≈ 1 billion TB), accounting for 99.35% of the total global traffic. This means that almost all internet data relies on data centers for processing, highlighting their importance.
According to a more popular saying nowadays, data centers are as vital as water plants and power plants, serving as the driving engine of the digital economy and the foundational support for national and social development.
█ Development Stages of Data Centers
Let’s first take a look at the history of data center development.
In the 1960s, humanity was still in the mainframe era. At that time, to store computer systems, storage systems, and power equipment, people built server rooms, which were referred to as “server farms”.
This “server farm” is considered the earliest prototype of a data center.
By the 1990s, with the birth and rapid development of the internet, many companies began to implement informatization. They built their own websites and set up numerous servers for email, FTP, OA (office automation), etc.
Some companies placed servers in internal company server rooms. Others, due to having few servers but not wanting to keep them in the office (due to noise, risk of power outages, and low security), opted to “host” them in the operator’s server room, renting space, power, and bandwidth from the operator, allowing them to manage and maintain the servers.
The early stage of data centers (Stage One)
Thus, the concept of data centers gradually began to take shape. In 1996, a US company called Exodus (specializing in server facility construction and bandwidth services) was the first to propose the term “IDC”.
This marks the early stage of IDC data center development.
In 1997, Apple launched a virtual machine software called “Virtual PC”. Later, VMWare also introduced the now-famous VMWare Workstation, marking the advent of the virtual machine era and laying the groundwork for the evolution of data centers.
As time progressed, the managed services of the first generation of data centers began to refine, extending from complete server hosting to website hosting, giving rise to virtual hosting services.
In other words, on a single server, through virtual hosting software, multiple website hosts can be virtually created and rented out to multiple clients.
In addition to websites, various services such as data storage space rental also emerged. This represents the second stage of IDC data centers.
The second stage of data centers
Moving forward to the early 21st century, companies like Amazon and Google introduced cloud computing, bringing data centers into the third stage (the cloud computing stage), which continues to this day.
The cloud computing stage is an upgrade evolution of the second stage. Through virtualization and container technologies, it has fully realized the pooling of computing resources in data center servers. All CPU, memory, hard disk, and other resources are managed by more powerful virtualization software and then allocated to users.
This evolution has transitioned from leasing physical hardware to leasing virtual hardware, and even software platforms and services. IaaS, PaaS, and SaaS have thus emerged.
The third stage of data centers (cloud computing stage)█ Structure of Data Centers
Next, let’s take a look at what parts make up a data center.
As mentioned earlier, a data center is a large server room. Thus, in terms of hardware types, data centers are quite similar to the internal server rooms of enterprises we often see, but with higher specifications, quality, and management levels.
Inside a data center
Overall, the hardware of a data center is divided into two categories: main equipment and supporting equipment.
Main equipment refers to devices that actually perform computing and communication functions, represented by IT computing devices such as servers and storage, as well as communication devices represented by switches, routers, and firewalls.
Supporting equipment exists to ensure the normal operation of main equipment, including various facilities.
Supporting facilities are further divided into various types, primarily including power distribution systems and cooling systems, as well as fire protection systems, monitoring systems, and building management systems.
-
Main Equipment
Let’s first look at the main equipment.
The most basic main equipment in a data center is, of course, the server. A server is essentially a high-performance computer, and everyone should have seen one. Inside, it contains components like CPU, memory, motherboard, hard disk, graphics card (GPU), and power supply.
Front view of a specific server model
Previously, servers were mostly based on Intel architecture (earlier, there were PowerPC, SPARC, etc.). Now, with changes in national policies, domestic CPUs have risen to occupy an increasing market share.These domestic CPUs use ARM architecture, providing better cost performance and lower costs.
Huawei chip Hi1620 (ARM architecture processor for data centers)
Servers are typically placed on racks (also called cabinets).
Server rack
A common standard rack typically has a height dimension of 42U. U is a unit that represents the external dimensions of the server, which is an abbreviation for unit, where 1U equals 4.445cm. The rack width can be either 600mm or 800mm.
The rack depth comes in various sizes, including 600mm, 800mm, 900mm, 1000mm, and 1200mm. Generally, IT equipment (servers) have deeper racks (1100mm or 1200mm), while communication equipment has shallower racks (600mm).
Comparison of rack depths for IT and communication equipment
Inside the rack, in addition to servers, there are also specialized storage devices such as disk arrays.
Nowadays, with the rise of big data, the amount of data generated by humanity is increasing every year, which raises the demand for the quantity and performance of storage devices.
Everyone should know that the mainstream computer storage drives are divided into HDD and SSD. HDD refers to traditional mechanical hard drives, while SSD refers to solid-state drives that are gradually becoming more popular.
SSD belongs to semiconductor storage, offering fast storage speeds and compact sizes, making it very popular. However, it is expensive. For data centers, due to cost-performance considerations, HDD remains the mainstream choice, while SSD is mainly used for high-end clients and high-performance demand businesses.
In addition to IT computing devices, there are also data communication devices such as switches, routers, and firewalls.
When talking about switches, we must mention a term—TOR, or Top of Rack.
TOR switches are common in the data center field. As the name suggests, they are switches located at the top of the rack. These switches are the lowest level of network switching devices in data centers, responsible for connecting the servers within the rack and linking them to higher-level switches.
Position of the TOR switch
In fact, rack switches do not necessarily have to be placed at the top of the rack. They can be located at the top, middle, or bottom of the rack. The reason they are usually placed at the top is simply because it is most beneficial for internal wiring.Above the rack is a row of racks, several rows of racks. To connect these racks and servers, data center networking technology is required.
The most popular networking architecture in data centers today is the Spine-Leaf network. I have an article (link) that specifically introduces this, so I won’t elaborate here.
Spine-Leaf network architectureIt is worth mentioning that, in order to transmit data with high bandwidth, data centers commonly use optical fiber instead of network cables. Therefore, optical fibers, optical modules, and optical communication devices (OTN, etc.) have become important components of data centers.Especially optical modules, high-speed optical modules (e.g., 400G) are very expensive and account for a significant portion of the data center’s costs, limiting development.
Optical moduleThere is also a popular optical communication term called DCI, which stands for Data Center Inter-connect. With the rise of distributed deployments, the data traffic between data centers is substantial, leading to high bandwidth requirements.Thus, operators and cloud service providers are developing DCI to build dedicated optical communication backbones between data centers, representing a significant market.The “East Data West Calculation” initiative in our country involves the interconnection of data centers, which has a significant stimulating effect on the DCI-related market.
Optical communication backbone network equipment
-
Supporting Equipment
Next, let’s look at the supporting and infrastructure devices and facilities of data centers.
First, let’s look at power distribution.
Power supply is the foundation for the normal operation of data centers. Without power, a data center is just scrap metal.
The main function of data center power distribution equipment is to control, protect, and manage the flow of electricity. The primary power distribution device is the distribution cabinet.
Data center distribution cabinets are divided into medium voltage and low voltage distribution cabinets. Medium voltage distribution cabinets typically operate at a voltage level of 10kV, connecting to the municipal power supply and feeding down to low voltage distribution cabinets. Low voltage distribution cabinets typically operate at a voltage level of 400V, further converting, distributing, controlling, protecting, and monitoring electrical energy.
Illustration of data center power distribution (case study)
Low voltage distribution cabinet in a data center (image from the internet)
In addition to distribution cabinets, to ensure normal power supply in emergencies, data centers are also equipped with a large number of UPS (Uninterruptible Power Supply) and even diesel generator sets.
“UPS + municipal power” is the traditional power supply solution. Nowadays, the more popular solution is “HVDC + municipal power.”
HVDC stands for High Voltage Direct Current. The differences between it and UPS involve more complex high-voltage knowledge, which I will have the opportunity to explain in detail later.
In short, the reliability and safety of “HVDC + municipal power” are higher, and the power supply efficiency surpasses that of “UPS + municipal power”, making it the mainstream development trend of uninterruptible power sources.
Let’s briefly discuss -48V and 220V.
Those with practical experience in the ICT industry know that IT devices like servers typically use 220V AC, while core network and wireless communication devices often use -48V DC.
Municipal power supply is generally AC. Data centers typically provide both -48V DC and 220V AC (through AC-DC conversion and DC-AC inversion).
In fact, DC is becoming the choice for more data centers (e.g., Google) as it has lower losses, higher energy utilization, and aligns with the high energy consumption trends of data centers under high computing power.
Now let’s take a look at cooling systems..
The cooling system is the second largest energy consumer in a data center, aside from the main equipment. I have detailed the power consumption data of data centers in my article on “East Data West Calculation”, so I won’t elaborate further here.
Currently, data center cooling mainly includes two methods: air cooling and liquid cooling.
Air cooling typically uses air conditioning systems similar to those in our homes. Like household air conditioners, data center air conditioning systems also consist of indoor and outdoor units. Relatively speaking, this technology is mature, the structure is simple, and maintenance is easy.
Liquid cooling uses liquid as a refrigerant for cooling.
Liquid has 25 times the thermal conductivity of air, and at the same volume, the heat removed by liquid is nearly 3000 times that of air. From a noise perspective, under the same cooling level, liquid cooling reduces noise by 20-35 decibels compared to air cooling. In terms of energy consumption, liquid cooling saves 30%-50% more electricity than air cooling.
Currently, liquid cooling technology is widely regarded as promising, but it is still in the exploratory stage. Overall, the market prospects for liquid cooling are very broad, with claims that the market size exceeds hundreds of billions.
Regarding cooling and heat dissipation, it is worth mentioning that cabinet pool-level, row-level, and cabinet-level cooling methods are on the rise, becoming the mainstream choice for newly built data centers.
Traditional cooling is room-level, cooling the entire server room. This method has a long cooling path, low efficiency, and cannot meet the heat dissipation needs of high-power devices, resulting in high energy consumption.
Cabinet pool-level, row-level, and cabinet-level cooling are designed around a cabinet pool, a row of cabinets, or a single cabinet.
Row-level cooling focuses on a row of cabinets for airflow design.
This method significantly shortens airflow paths, resulting in high cooling efficiency.
In addition to power distribution and cooling, data centers also have some devices and facilities related to management and operation, such as dynamic environment monitoring systems, building automation systems, and fire protection systems.
Dynamic environment monitoring refers to the monitoring and management of the operational status of data centers in real-time.
In recent years, based on traditional dynamic environment monitoring systems, DCIM has begun to evolve.
DCIM stands for Data Center Infrastructure Management, a term proposed by the well-known consulting firm Gartner. Its management scope is more comprehensive, using tools to monitor, manage, and control all IT main equipment and supporting infrastructure in data centers.
The fire protection system in data centers is quite interesting. Since the server rooms are filled with electronic devices, in the event of a fire, direct spraying of water, foam, or powder is not possible.
So what to do? Gas fire extinguishing.
When a fire occurs, the fire and smoke sensors sound an alarm, and then inert gases like argon or nitrogen can be released in the server room area, depriving the flames of oxygen and extinguishing the fire (which can be done in about a few seconds).
Gas fire extinguishing (image from the internet)
-
Modular Data Centers
Data centers are a massive system, and the construction process is very complex and lengthy. In recent years, to deploy data centers more quickly and flexibly, manufacturers have introduced the concept of modular data centers.
Simply put, this involves integrating the data center’s structural systems, power distribution systems, HVAC systems, fire protection systems, lighting systems, and cabling into modular “building blocks”. Then, these “blocks” can be transported to the site for simple installation and setup, enabling quick construction and deployment.
Using this approach, the construction cycle for large data centers has been reduced from 18-24 months to about 6 months, resulting in significant economic benefits.
Modular data center (model)█ Conclusion
That concludes the introduction to data centers.
As I mentioned earlier, data centers are important infrastructures in the digital age, serving as crucial carriers of computing power that directly determine a country’s digital competitiveness.
Driven by the “East Data West Calculation” strategy, domestic data centers are set to experience another wave of development. According to statistical data, the market size of data center services in China is expected to reach 320.05 billion yuan in 2022, with an average compound annual growth rate of 27.0%. By 2025, domestic IT investment in data centers is expected to reach 707.09 billion yuan.
In addition to the explosive growth in quantity, data centers are also moving towards greener and smarter directions, actively incorporating AI to improve energy efficiency and reduce operational complexity.
In the future, whether data centers will undergo new forms of change remains to be seen!
References:
1. “Data Center Industry Map Research Report”, China Academy of Information and Communications Technology
2. “Data Center Energy White Paper”, Huawei
3. “Data Center Special Report”, Dongguan Securities
——— More exciting videos can be found on Huawei Training video account ———
Don’t forget to “collect + like + watch” all three! 👇