What is Step Functions?

Orchestrate tasks – You can create workflows that orchestrate a series of tasks, or steps, in a specific order. For example, Task A might be a Lambda function which provides inputs for another Lambda function in Task B. The last step in your workflow provides the final result.

Choose tasks based on data – Using a Choice state, you can have Step Functions make decisions based on the state’s input. For example, imagine that a customer requests a credit limit increase. If the request is more than your customer’s pre-approved credit limit, you can have Step Functions send your customer's request to a manager for sign-off. If the request is less than your customer’s pre-approved credit limit, you can have Step Functions approve the request automatically.

Error handling (Retry / Catch) – You can retry failed tasks, or catch failed tasks and automatically run alternative steps.

For example, after a customer requests a username, perhaps the first call to your validation service fails, so your workflow may retry the request. When the second request is successful, the workflow can proceed.

Or, perhaps the customer requested a username that is invalid or unavailable, a Catch statement could lead to a Step Functions workflow step that suggests an alternative username.

For examples of Retry and Catch, see Handling errors in Step Functions workflows.

Human in the loop – Step Functions can include human approval steps in the workflow. For example, imagine a banking customer attempts to send funds to a friend. With a callback and a task token, you can have Step Functions wait until the customers friend confirms the transfer, and then Step Functions will continue the workflow to notify the banking customer that the transfer has completed.

For an example, see Create a callback pattern example with Amazon SQS, Amazon SNS, and Lambda.

Process data in parallel steps – Using a Parallel state, Step Functions can process input data in parallel steps. For example, a customer might need to convert a video file into several display resolutions, so viewers can watch the video on multiple devices. Your workflow could send the original video file to several Lambda functions or use the optimized AWS Elemental MediaConvert integration to process a video into multiple display resolutions at the same time.

Dynamically process data elements – Using a Map state, Step Functions can run a set of workflow steps on each item in a dataset. The iterations run in parallel, which makes it possible to process a dataset quickly. For example, when your customer orders thirty items, your system needs to apply the same workflow to prepare each item for delivery. After all items have been gathered and packaged for delivery, the next step might be to quickly send your customer a confirmation email with tracking information.

For an example starter template, see Process data with a Map.