On-Demand Instances
- What it is
- Pay by the second (or hour) for compute, with no commitment and no upfront cost.
- Start and stop whenever; the price is the published list price.
- When to use
- Short-term, spiky, or unpredictable workloads.
- New applications being tested where usage is unknown.
- Workloads that cannot be interrupted and cannot commit to a term.
- Trade-off
- The most flexible option and the most expensive per hour.
- Not to be confused with
- Reserved Instances / Savings Plans — cheaper in exchange for a 1- or 3-year commitment.
- Spot Instances — far cheaper, but AWS can reclaim the capacity.