Companies that use serverless computing outsource their backend components. This includes databases, calculations, storage, or the processing of data flows. The user therefore concentrates on the design. Some people use the term Backend as a Service (BaaS).

This dynamic can be compared very simply to that of a large restaurant. You are responsible for the menu title, the décor, the atmosphere and the creation of the dishes and their presentation, but you delegate all the work in the kitchen. Another team takes care of peeling the potatoes, making sure the products are fresh and available, raising the sauce and cooking your entrecote or vegetarian steak to perfection. The icing on the cake is that you pay them by the hour and only when they are working.

Presented in this way, serverless computing makes you want to lick your lips and we understand that its use is bound to grow. The major constraint to bear in mind is that you are entering into a relationship of dependence with a third-party company. By delegating, you increase the range of possibilities, but you also lose some of your autonomy. Serverless computing also has its limitations when thinking about complex applications. Imagining an architecture based solely on serverless computing remains very risky and difficult to implement in 2020.

Serverless computing operates on an event driven basis, i.e. when an event occurs, a function is executed. It differs in this sense from the use of Virtual Machines (VMs) or PaaS models (Platform as a Service) for example. The designer or developer rents a kind of space to perform a function and only pays when the function is actually used. The underlying concept is therefore that the user does not pay for the server as such but for the service it offers. The developer must therefore create the application code and the associated stateless (or neutral) functions.

At present, serverless computing mainly concerns applications that are still very simple. The following examples can be cited:

• Integration of a chatbot;
• Real-time processing of structured or unstructured data;
• Automatic execution of tasks, especially for Machine to Machine, M2M communications;
• Availability of data in different formats.

These few illustrations represent only a small fraction of what serverless computing can achieve. In practice, serverless computing represents a form of evolution of the PaaS model and is ideally suited to any application with volatile or cyclical workloads. Serverless computing allows costs to be rationalised while the PaaS model remains more rigid. 

Many large companies are currently using serverless computing. The percentage is 40% according to a study conducted by O'Reilly, a publishing house specialising in IT. Well-known names such as Coca-Cola or Autodesk are often cited as examples of users. Netflix is another concrete example. The company uses AWS Lambda to adapt and resize its images in relation to the interface. This function is performed on demand, which avoids storing an astronomical number of media files.