On-Demand Instances allow us to pay only for the compute capacity we use. The instances are charged by the hour or by the second depending on the instance type. There are no long-term commitments with On-Demand Instances; we simply launch an instance when needed and terminate it when not required any more. On-Demand Instances are great for any task that has an unpredictable or temporary workload, such as the following:

• Processing tasks from a variable source such as a message queue
• Running code execution in response to an event
• Covering temporary burst capacity requirements for EC2 clusters
• Developing and testing applications
• PoC deployments