I've been digging into Flux's pricing page, and the core unit is this "task." It's not immediately obvious from a cloud cost perspective what that maps to in practice. The docs say "a single execution of a Flux workflow," but I need a concrete translation to understand the cost implications.
From what I can piece together, think of it like this:
* A **task** is essentially a single step in your data pipeline that runs on a Flux worker. It's not the entire workflow, but each unit of work within it.
* If you have a workflow that: fetches data → transforms it → loads it to a warehouse, that's likely **three tasks**.
* Crucially, a task seems to be tied to a compute run time. If your "transform" step is a heavy Python script running for 10 minutes, that's one task. A simple SQL query that finishes in 2 seconds is also one task.
The billing meter is "task compute hours." So if you run 1,000 tasks in a month, and each task runs for an average of 2 minutes, your total consumption is:
```plaintext
1000 tasks * (2 minutes / 60) = ~33.33 task compute hours
```
You'd then multiply that by the hourly rate per task type (Standard, Power, etc.).
The part I'm still trying to map is how this compares to, say, AWS Step Functions state transitions or Azure Logic App actions. It feels more granular. A key question for cost optimization: does a retry of a failed step count as a new task? Because if so, flaky data sources or API calls could directly spike your bill. Has anyone seen this detail in their usage reports?