第一章 云朵工厂：什么是云计算？ / Chapter 1 Cloud Factory: What is Cloud Computing?

Lily is a 10-year-old who loves playing *Egg仔 Party* (Eggy Party) on weekends. Last Saturday, she was about to win the championship when the game suddenly froze, the character stopped moving, and the screen showed a spinning loading icon. “It's probably the Cloud Factory being too busy,” Mom said, tapping her phone screen to show the network speed bar dropping to the bottom. “What's a Cloud Factory?” Lily blinked. “Is it a place that makes cotton clouds?” Mom laughed and opened a video of a huge data center on her phone: rows upon rows of black cabinets, with blinking lights like stars, and engineers in uniforms walking around checking equipment. “This is the real Cloud Factory. It's not a place that makes weather clouds, but a huge cluster of supercomputers thousands of kilometers away, connected to our phones and computers via the Internet. Just like a supermarket has shelves full of snacks for everyone to buy, the Cloud Factory has endless computing power, storage space, and network bandwidth for everyone to use at any time, as long as you are connected to the Internet.”

<!-- 标签：云计算 云朵工厂 核心组成 -->

云朵工厂的核心组成：藏在云里的超级设备

The Cloud Factory is not a single supercomputer, but a cluster of tens of thousands to millions of physical servers connected together, usually located in places with low temperatures and sufficient electricity (such as northern China's Zhangjiakou or Norway's Arctic Circle) to reduce cooling and power costs. According to the 2024 China Cloud Computing White Paper released by the China Academy of Information and Communications Technology, a standard public cloud data center (the real Cloud Factory) has a construction area equivalent to 10 football fields, with a total power consumption equivalent to the residential electricity usage of a small and medium-sized city[^2].

To make it easier for you to understand, we can compare the Cloud Factory to a large supermarket that only sells “computing services” instead of snacks. It has four core functional areas: 1. Super Computing Cluster (计算集群): Equivalent to the supermarket's checkout counters and packaging staff. Each server in the cluster has a computing power hundreds of times higher than an ordinary home computer, responsible for processing all kinds of tasks: calculating your game character's movement trajectory, decoding the video you are watching, and even helping doctors analyze medical images. For example, the supercomputing cluster of Alibaba Cloud can process more than 10 billion transaction requests per second during the Double 11 Shopping Festival, which is equivalent to processing the entire population of China's order data in 1 second. 2. Storage Resource Pool (存储资源池): Equivalent to the supermarket's warehouse. It uses a large number of hard drives and solid-state drives to store all kinds of data: your game progress, photos and videos taken with your phone, online course videos, and even the code of the software you use. The storage capacity of a medium-sized Cloud Factory can reach the EB level (1EB = 10 million GB), which is enough to store 2 billion hours of 4K video, or all the photos taken by all people in the world for 10 years. 3. Network Bandwidth Pool (带宽资源池): Equivalent to the supermarket's highway and delivery trucks. It is responsible for transmitting data between your device and the Cloud Factory at high speed and low latency. The larger the bandwidth, the faster the data transmission, just like a wider highway can allow more cars to pass through at the same time. Cloud vendors usually lay optical fibers around the world to ensure that users in different regions can access the nearest Cloud Factory node, reducing transmission delay. For example, if you are in Beijing, you will connect to the Beijing node of the Cloud Factory, and if you are in Guangzhou, you will connect to the Guangzhou node, so the speed will be faster. 4. Operation and Maintenance Team (运维团队): Equivalent to the supermarket's security and cleaning staff. They work 24 hours a day to monitor the operation of all equipment in the Cloud Factory, replace broken hard drives in time, upgrade software, and deal with network attacks, ensuring that the Cloud Factory can run stably without downtime.

<!-- 标签：虚拟化 核心技术 -->

核心技术：虚拟化（Virtualization）：把大蛋糕切成小份

You may have a question: there are so many users using the Cloud Factory, will the resources be not enough? This is the credit of the core technology of cloud computing: Virtualization (虚拟化).

Simply put, virtualization is like cutting a large birthday cake into small pieces for every guest. The physical server in the Cloud Factory is the large cake, and the virtualization layer (called Hypervisor) is the knife that cuts the cake. It splits the CPU, memory, and storage of a single physical server into dozens or even hundreds of independent virtual servers. Each user gets a “piece” of virtual resources, and they think they are using an independent physical server, but in fact they are sharing the same physical hardware with other users, and they will not affect each other.

Without virtualization technology, cloud computing would not be able to achieve low cost and high resource utilization. For example, if there is no virtualization, a physical server with 32 cores and

Without virtualization technology, cloud computing would not be able to achieve low cost and high resource utilization. For example, if there is no virtualization, a physical server with 32 cores and 64GB of memory can only run one application at most. If the application only needs 4 cores and 8GB of memory, the remaining 28 cores and 56GB of memory will be completely wasted. According to statistics from the International Data Corporation (IDC), before virtualization technology was popularized, the average resource utilization rate of physical servers in enterprises was less than 15%, which means that more than 85% of the computing power you paid for was lying idle[^1]. After virtualization technology is applied, this 32-core physical server can be split into 8 independent virtual servers, each with 4 cores and 8GB of memory. Each virtual server runs its own application without interfering with each other, and the overall resource utilization rate of the physical server can reach more than 70%. This is like a large pizza cut into 8 small pieces, each person gets a piece, and no food is wasted. The virtualization layer that completes this splitting work is called Hypervisor (虚拟机监视器), which is the "kit