Cloud computing

Cloud computing, usually just called the cloud, means renting computing power, storage and software over the internet from a provider that owns and runs the servers, and paying only for what you use. The customer does not buy servers or keep them in a building of its own. It can rent a server in minutes and stop paying for it the moment it no longer needs it.

Here is how it works in general. The cloud provider buys hundreds of thousands of servers and runs them in large data centres in many cities. Several providers have data centres in Mumbai, Hyderabad, Pune and Delhi. Software splits each physical server into many virtual machines, each of which behaves like a separate computer. A customer logs in to the provider's website, or uses its APIs, and asks for what it needs: a virtual server, storage or a database. The provider sets it up within minutes. The customer's app runs on those rented machines. At the end of the month, the provider bills the customer for the hours of computing and the gigabytes of storage it used. Many customers share the same physical machines, without being able to see each other's data. This is called multi-tenancy.

Customers can rent at three levels. They can rent bare virtual servers and run everything on them. They can rent ready-made services, such as a database that the provider installs, updates and backs up. Or they can rent a finished application, which is called software as a service.

Take, for example, three students who build an exam-practice app. On a developer cloud, they rent one virtual server and a managed database, and rent more servers for the weeks before board exams, when use rises.

Consider a CPAP machine, which helps a patient breathe at night. The CPAP machine sends each night's data to the maker's cloud, where the patient's doctor can see it the next morning.

Say an AI start-up needs GPUs to train a model for two weeks. It rents them through a GPU cloud and pays by the hour, instead of buying chips that cost crores.

How is the cloud different from on-premises computing? On-premises means the company buys its own servers, keeps them in its own building and runs them with its own staff. Many large companies use both, which is called hybrid cloud. In the cloud, the provider secures the buildings and the machines, while the customer must still secure its own data and settings. This split is explained in the shared responsibility model.