What is AWS Lambda?

The Lambda function model

A Lambda function combines code with configuration. The code exposes a handler that receives an event and invocation context. The configuration defines details such as the runtime, memory, timeout, environment variables, networking, concurrency controls, and an IAM execution role.

Functions can be invoked directly or connected to event sources. HTTP requests, object changes, queue messages, stream records, schedules, and events from other AWS services can all initiate work through different integration models.

What happens during an invocation

Lambda receives an invocation, selects or creates an execution environment, initializes the runtime and your code when needed, and calls the handler. An available environment may be reused for later invocations, so initialization work can be reused but must never be treated as durable application state.

The caller and integration determine how success, errors, retries, and duplicate delivery behave. A synchronous API request and a batch of queue messages have different failure contracts. Design the function around the contract of its event source rather than treating every invocation the same.

When Lambda is a good fit

Lambda is a strong option for event handlers, APIs, automation, file processing, stream or queue consumers, scheduled jobs, and workflow tasks that fit its execution model. It works especially well when demand varies and each unit of work has a clear input, output, and ownership boundary.

A function should not become the default answer to every compute problem. Review execution duration, concurrency behavior, startup sensitivity, networking, package size, local tooling, data gravity, and cost shape before choosing the service.

A production Lambda checklist

Treat retries and duplicate events as normal. Make handlers idempotent where the integration can redeliver work, validate every event at the boundary, and record enough context to trace one business operation across services.

Grant the execution role only the actions and resources the function needs. Set explicit timeouts and concurrency controls, define failure destinations or dead-letter handling where appropriate, protect secrets, and alarm on user-visible failure rather than invocation count alone.