SysPad › Lessons › orchestration lessons › AWS Step Functions

How AWS Step Functions works

Serverless workflow orchestration. Coordinates Lambda, ECS, and other services as a state machine. Standard workflows are durable and exactly-once; Express workflows trade durability for high throughput.

As soon as a process has more than a couple of steps, "Lambda calls Lambda calls Lambda" turns into a tangle of retries, error handling, and state nobody can follow. Step Functions pulls that logic OUT of your code into a visual state machine: it runs each step, remembers where you are, and handles failures, so your functions just do one job each.

What you will learn

What problem does orchestration solve that chaining Lambdas directly creates?
What are states, and how do branching, parallelism, and iteration work?
How do built-in retries and catch replace hand-written error handling?
Standard vs Express workflows: which for which job?

Logic out of code, into a workflow

A Step Functions state machine is a workflow of states. Each state does one thing, often invoking a Task (a Lambda, an ECS run, another AWS API), then passes its output to the next state. Step Functions tracks position and data for you.

States: branch, parallel, iterate

Beyond Task states, there are Choice (branch on data), Parallel (run branches at once), Map (run a step over each item in a list), and Wait (pause). You compose these into the exact flow you need.

Crucially, the orchestration is declarative, the diagram is the logic, which makes complex processes visible and auditable.

⚓ Order processing, ETL pipelines, ML pipelines, human-approval flows, and any multi-step saga across services.

Built-in retries and catch

Each state can declare Retry (with backoff) and Catch (route to a handler state) for failures. You configure resilience instead of writing retry loops and try/catch glue in every function.

Standard vs Express workflows

Standard workflows are durable and long-running (up to a year), with exactly-once execution and full execution history, ideal for business processes and human-in-the-loop. Express workflows are for very high-volume, short-duration event processing (capped at 5 minutes per execution), cheaper per run, but with at-least-once semantics and limited history.

Flashcards

What does Step Functions move out of your Lambda code?: The orchestration logic: sequencing, branching, state, and error handling. Functions do one job; the state machine coordinates them.
Name the state types for branching, concurrency, and iteration.: Choice (branch), Parallel (concurrent branches), Map (iterate over a list); plus Task and Wait.
How do you handle a flaky step without writing retry code?: Declare Retry (with backoff) and Catch on the state; resilience becomes workflow config, not code.
Standard vs Express workflows, the headline difference?: Standard: durable, long-running (up to 1 year), exactly-once, full history. Express: high-volume, short, cheaper, at-least-once, limited history.
Why orchestrate instead of chaining Lambdas that call each other?: Direct chaining scatters retries, error handling, and state everywhere. A state machine centralises and visualises it, making complex flows reliable and auditable.

Sources

Open the SysPad canvas · Official AWS Step Functions documentation