Reserved Instances and Savings Plans#
Companies can spend a lot of money on cloud servers, and part of building on AWS is optimizing your spend and getting the most out of it.
Initially, AWS allowed you to “reserve” an instance. This meant:
- You commit to paying for at least a certain amount of compute on a specific EC2 instance type, for a period of 1-3 years.
- In return for your commitment, AWS gives you a significant discount on that compute.
It’s like buying in bulk, but it comes with constraints. The reservation is tied to a specific instance type and region (sometimes even a specific availability zone, or AZ).
More recently, AWS has introduced Savings Plans, which offer more flexibility while still giving customers similar discounts. In brief:
- You commit to 1-3 years of usage.
- You can change the AZ of your instances (e.g., from your
public-asubnet topublic-b). - You can change the instance type, within a “family” (e.g., the
t3int3.smallandt3.medium).
Savings can range from 26% to 72%! In many cases, turning off a server every night after business hours saves less money than just adding a savings plan and running the thing 24/7. If you’re sure you’re going to run a server for a year, you’re wasting money by not having a savings plan.
Next#
If you want to save even more, see spot-instances - unused AWS capacity at a ~90% discount, with the caveat that AWS can reclaim it at any time.