AWS Step Functions: Workflow Orchestration

While SQS, SNS, and EventBridge excel at decoupled, event-driven communication, some workflows require coordination — steps that must happen in a specific sequence, with branching logic, error handling, and retries. AWS Step Functions is a serverless workflow orchestration service that lets you coordinate multiple AWS services into structured, visual workflows.

---

Choreography vs Orchestration

Event-driven systems use two coordination styles:

Choreography (Decentralised)

Each service reacts to events independently. There is no central coordinator — services communicate through events.

[Order Service] ——event——> [Payment Service] ——event——> [Shipping Service]

Pros: Loose coupling, each service is independent. Cons: Hard to see the full workflow, difficult to handle failures across services.

Orchestration (Centralised)

A central coordinator (Step Functions) manages the workflow. It calls each service in the right order, handles errors, and tracks state.

[Step Functions Orchestrator]
     |——> Step 1: Validate Order
     |——> Step 2: Process Payment
     |——> Step 3: Reserve Inventory
     |——> Step 4: Ship Order
     |——> Step 5: Send Confirmation

Pros: Clear visibility, centralised error handling, easy to modify workflow. Cons: Central coordinator introduces a dependency.

When to use orchestration: Use Step Functions when the workflow has sequential steps, branching logic, error handling requirements, or when you need visibility into the workflow state.

---

How Step Functions Works

State Machines

A Step Functions workflow is called a state machine. It is defined using Amazon States Language (ASL), a JSON-based language that describes the states, transitions, and logic of the workflow.

States

Each step in the workflow is a state. Step Functions supports several state types:

State Type	Purpose	Example
Task	Do work (invoke Lambda, call API, run ECS task)	Process payment
Choice	Branch based on a condition	If amount > 100, require approval
Wait	Pause for a duration or until a timestamp	Wait 24 hours before sending reminder
Parallel	Execute branches simultaneously	Check credit and check inventory at same time
Map	Process each item in a collection	Process each line item in an order
Pass	Pass input to output (useful for transformations)	Restructure data
Succeed	Mark the workflow as successful	End of happy path
Fail	Mark the workflow as failed	End with error

---

Workflow Types

Standard Workflows

• Duration: up to 1 year
• Execution model: exactly-once
• Pricing: per state transition
• Use case: long-running, business-critical workflows
• State history: full execution history retained for 90 days

Express Workflows

• Duration: up to 5 minutes
• Execution model: at-least-once (synchronous) or at-most-once (asynchronous)
• Pricing: per execution and duration
• Use case: high-volume, short-duration event processing
• State history: sent to CloudWatch Logs

Feature	Standard	Express
Max duration	1 year	5 minutes
Execution guarantee	Exactly-once	At-least-once / at-most-once
Pricing	Per state transition ($0.025 per 1,000)	Per execution + duration
Throughput	Up to 2,000/s start rate	Over 100,000/s start rate
Best for	Order processing, ETL, human approval	IoT data processing, real-time transforms

---

Amazon States Language (ASL) Example

Here is a simple order processing workflow: