What does serverless mean? Somebody still runs the servers

Serverless is about who manages what

Imagine a small service that resizes a photo when someone uploads it. One approach is to rent a computer, install the application, and keep that computer ready. Another is to supply a function that a managed platform runs when an event arrives.

In the second approach, computers still do the work. “Serverless” does not mean “without servers.” It describes moving more execution-infrastructure management to the provider. You focus on the function and its surrounding application rather than operating the underlying host yourself.

AWS Lambda’s documentation describes a managed way to run code without provisioning and managing servers. This guide uses a function-style mental model, not every product marketed as serverless.

Try it: quiet morning, busy afternoon

The fixed allocation below has four pretend job slots. Move the slider above four: some jobs must wait in this simplified picture. Then move it to zero: the allocation still exists, even though no jobs are active.

The managed side grows with each job in the illustration. Its instant growth is a teaching shortcut. Real systems have concurrency limits, queues, startup time, and failure handling. This is not a capacity forecast or a cloud bill calculator.

You still have important jobs

Someone must decide what the function is allowed to read, where it stores output, and how failures are reported. Those decisions do not vanish when a provider runs the computers.

For the photo service, ask: which upload starts the function, where is the resized image written, and what happens if processing runs twice? Keeping a record and making repeated work safe can matter more than the choice of hosting label.

A temporary execution environment is not a dependable permanent filing cabinet. Save durable information in an appropriate storage service. Reusing an environment can improve performance, but a function should not depend on its previous local state always being present.

When it helps—and when to pause

Event-driven, intermittent work can be a useful fit because you do not choose and run a host for every function yourself. But a continuously busy application may need a different cost and architecture comparison. Managed does not mean unlimited.

Zero pretend jobs in our slider does not guarantee a zero total bill. Storage, databases, logging, networking, minimum charges, or other application components can still cost money. Compare the actual services and pricing for the deployment you intend to use.

For a beginner project, first describe the task in one sentence: “When a photo arrives, make a smaller copy.” Then identify the input, output, permission, and failure case. If a static file can solve the problem, you may not need a function at all.

This site’s learning exercises run in your browser. Running a cloud function for every slider movement would add complexity without helping you understand this concept.

Keep the labels in perspective

Serverless and microservices answer different questions. A microservice describes how an application is split into services; serverless concerns aspects of how code runs and infrastructure is managed. One does not require the other.

The older serverless and microservice architecture guide explores those labels. For the journey that gets a request to an application, start with DNS and website loading.