It is not in the sky
The word “cloud” suggests something weightless and floating. In reality, cloud computing means using computers that are located somewhere else, reached over the internet, instead of using only the computer in front of you. The name comes from the habit of drawing the internet as a cloud-shaped blob in network diagrams, a way of saying “some network stuff happens here, and we don’t need the details.”
So when a photo is “in the cloud,” it is stored on hard drives or flash drives inside a building full of machines, owned by a company, in a real place on the planet.
Data centers
Those buildings are called data centers. They hold thousands of servers, which are computers designed to run continuously and answer requests from other computers. A data center also has fast network connections, backup power supplies, and large cooling systems, since computers turn nearly all the electricity they use into heat. Data centers use a significant amount of electricity, which ties the cloud directly to the energy sources that power the grid.
Providers typically spread their equipment across many data centers in different regions. Copies of data may be kept in more than one place, so that a failure in one machine or building need not mean lost data.
Renting instead of owning
The central idea of cloud computing is renting computing as a service. Traditionally a business that wanted to run a website or store records had to buy servers, house them, power them and maintain them. With the cloud, it pays a provider for the resources it uses, and can typically add or remove capacity quickly. A small project might use a sliver of a machine, while a large one uses thousands.
A technique called virtualization makes this practical. Software divides one physical server into several virtual machines, each acting like a separate computer with its own operating system, while sharing the same hardware. Customers are kept isolated from each other by design. Newer approaches, such as containers, package programs in lighter-weight ways, but the sharing idea is similar.
Common kinds of cloud service
People often group services into broad layers.
- Infrastructure: raw computing power, storage and networking that a customer configures themselves.
- Platform: a ready-made environment where developers can run their programs without managing the underlying machines.
- Software: complete applications used through a browser or an app, such as email, document editors or video calls, where the provider handles everything behind the scenes.
Most people use the last kind every day without thinking about it. Streaming a video, syncing notes between a phone and a laptop, and using online maps all rely on remote servers doing the heavy work and sending back the result.
How a request travels
When you open a cloud app, your device sends a request across the network, following the pathways described in how the internet works. The request reaches a server, which may fetch your data from storage, run some calculations and send back a response. Your own device mostly displays results. This is why some web apps work on a modest phone: the demanding work happens elsewhere.
It also explains why the cloud depends on connectivity. Without a network connection, many cloud tools become limited or stop working, unless they keep a local copy of your data.
Why it is popular
- Flexibility: capacity can grow or shrink with demand, so a business does not need to buy equipment for its busiest day.
- Access: data and programs can be reached from many devices.
- Shared expertise: large providers can invest in maintenance, backups and reliability that small organizations may find costly.
- Large-scale tasks: big jobs, such as training a machine learning system on huge amounts of data, can use many machines at the same time, and many artificial intelligence services run this way.
Trade-offs
The cloud is not automatically better. There are real considerations.
- Dependence: if a provider has an outage, many services that rely on it can stop at once.
- Control and privacy: your data sits on someone else’s equipment, so you depend on the provider’s practices and on the settings you choose. Data may be protected with encryption basics such as scrambling it in storage and in transit, but protections vary between services.
- Costs: pay-as-you-go pricing can be cheap for small or uneven needs and become expensive for large, steady ones. Which is cheaper depends on the situation.
- Law and location: where data physically sits can matter for legal rules, and these differ between countries.
- Lock-in: moving a large system from one provider to another can be difficult.
A useful way to think about it
The cloud is best understood as electricity-like: a utility. Most people do not run their own power plant but plug into a shared supply and pay for what they use. Cloud computing offers something similar for processing and storage. The convenience is real, and so is the fact that someone else’s machines are doing the work, which is worth remembering when deciding what to put there.