On-Demand Instances is a pricing model of AWS that allows you to pay for compute capacity billed per hour or second without committing to long-term use. You can equip On-Demand Instances instantly and use them for a specific duration without any upfront payment. With On-Demand instances, you can decide when to launch, stop, hibernate, start, reboot, or terminate instances.
If you are looking for an instance type where you can run short-term workloads, signing a long-term (1-3 years) contract to use Reserved Instances or Saving Plans is not what you need. Look into On-Demand instances and see how it is suitable for you.
With On-Demand Instances, you:
On-demand instances are recommended for organizations that prefer flexible access to computer resources without upfront payment or long-term commitment.
As mentioned above, Spot Instances involves bidding on an unused compute capacity. Unlike On-Demand instances, Spot instances are not always available. You have to request and specify an amount you are willing to pay per hour per instance, and if your bid is higher than the current spot price, your instances start running. If another bid exceeds yours, your instances will be terminated with short notice.
Choosing between On-Demand and Spot Instances requires a lot of consideration of some factors: characteristics, costs, and business objectives. The type of instances that you should use depends on what your workload is. For example, if your workloads need to run continuously, On-Demand instances are much more suitable.
When it comes to costs, Spot Instances offer more cost savings (up to 90% off), but On-Demand instances are more stable and predictable. And if you have short-term projects, spot instances are a better option because of their lower costs. Long-term projects require stability and predictability.
While On-Demand instances and Reserved Instances share similarities, they differ in application. Reserved Instances (RI) is a pricing model that offers a discount if you commit to using a fixed amount of compute capacity per hour for a one- or three-year period. This commitment means you agree to pay for the instance upfront (all upfront, partial upfront, or no upfront), which reduces your overall cost compared to paying for On-Demand instances on an hourly basis. You can save up to 72% on the On-Demand hourly rate if you use Reserved Instances.
Here are some examples of when to use On-Demand Instances:
On-demand instances are instances that you only use on demand. With on-demand instances, you only pay for the time that your instances are running. The price per second for on-demand instances is fixed, and no need to sign any long-term commitment (1-3 years), which makes it suitable for irregular workloads that cannot be interrupted. Despite its flexibility, On-Demand Instances are more expensive and may sometimes result in unexpected costs. A tool like Octo can help you understand and optimize your On-Demand Instance costs.
