On-Demand instance - This is used where you need flexibility.
- So if you need the low cost and flexibility of EC2 without any upfront payment or long-term commitment.
- Typical use case so this is where you’re building a site from scratch, and you just want to see what the application looks like, and you’re coding it for the very first time.
- It’s also good and useful for short term.
- So applications with short term or spiky or unpredictable workloads that can’t be interrupted.
Reserved Instances - Essentially it’s the opposite of that.
- So it’s where you’ve got predictable usage.
- This is where you understand the usage patterns of your applications, and it’s a steady state predictability behind it. It’s also where you’ve got specific capacity requirements.
- It’s also very useful way you can afford to pay upfront maybe you’ve gone out and gotten VC funding,
- so you’re able to burn some cash for a couple of years.
- You can make some upfront payments to reduce the total computing costs even further.
Convertible Reserved Instances.
- So these are similar to Standard Reserved Instances except you have the option to change to a different reserved instance type with equal or greater value.
- You have compute-based EC2 instances.
- You might have GPU-based EC2 instances.
- And the cool thing about Convertible Reserved Instances, it allows you to change between the different classes of EC2 instances.
Scheduled Reserved Instances
- They basically allow you to launch within the time window that you find, so you can match your capacity reservation to a predictable recurring schedule that only requiresa fraction of the day, week, or month.
- They operate at a regional level.