Compute Savings Plans and Reserved Instances (RIs) are both ways to save money on EC2, but they operate on fundamentally different principles, and the most surprising thing is how much flexibility Compute Savings Plans offer without sacrificing significant savings.
Let’s see them in action. Imagine you have a fleet of t3.medium instances running in us-east-1a and us-east-1b.
Reserved Instances (RIs)
With RIs, you commit to a specific instance family, region, OS, and tenancy for a 1 or 3-year term.
- Standard RIs: Offer the most flexibility within a specific instance family and region. If you buy a
t3.mediumStandard RI inus-east-1, you can use it for anyt3.mediuminus-east-1, regardless of AZ, OS, or tenancy. - Convertible RIs: Offer even more flexibility, allowing you to change instance attributes like family, OS, and tenancy, but with a lower discount than Standard RIs.
If you have a predictable workload that consistently uses specific instance types in particular regions, RIs can be a good fit. For example, if you know you’ll always need 10 m5.large instances in us-west-2 for the next three years.
Compute Savings Plans (CSPs)
CSPs are a more flexible pricing model. Instead of committing to specific instance attributes, you commit to a dollar amount of usage per hour.
- Compute Savings Plans: Offer the most flexibility. You commit to a dollar amount per hour for compute usage. This commitment applies to EC2 instances, Fargate, and Lambda. You get discounts regardless of instance family, region, OS, tenancy, or even compute service.
- EC2 Instance Savings Plans: Offer a more targeted discount, but still more flexible than RIs. You commit to a dollar amount per hour for EC2 instance usage. The discount applies across instance families, regions, OS, and tenancy within EC2.
Let’s say you have a mix of t3.medium, m5.large, and c5.xlarge instances across different regions. With Compute Savings Plans, you can commit to spending $10/hour on compute. AWS will automatically apply the discount to whichever instances are running and eligible, maximizing your savings without you having to micro-manage instance types or regions.
The core problem both RIs and CSPs solve is the difference in price between On-Demand and committed usage. On-Demand pricing is for when you need immediate capacity without commitment, but it’s the most expensive. By committing to a certain level of usage, you signal to AWS that they can provision capacity more efficiently, and they pass those savings on to you.
How they work internally:
When you purchase an RI, AWS earmarks capacity for you. If you have a t3.medium RI in us-east-1, AWS ensures there’s a t3.medium available for you in that region. If you’re not using it, that capacity might sit idle, but you’ve paid for it.
With Compute Savings Plans, it’s less about earmarking specific capacity and more about a monetary commitment. AWS tracks your hourly compute spend. If you’re under your committed dollar amount, the discount is applied. If you exceed it, you pay the On-Demand rate for the excess. The system dynamically matches your eligible usage against your commitment to give you the best possible discount.
The levers you control:
- Commitment Term: You can choose between a 1-year or 3-year commitment. Longer terms offer higher discounts.
- Commitment Amount (for CSPs): You decide how much you want to commit to spending per hour. This is a critical lever for balancing savings and flexibility.
- Commitment Type (for EC2 Instance Savings Plans): You can choose between "Compute Savings Plans" (most flexible) or "EC2 Instance Savings Plans" (slightly less flexible but potentially higher discount on EC2).
- Coverage: You can choose to cover "All compute services" (for Compute Savings Plans) or just "EC2 instances" (for EC2 Instance Savings Plans).
The most surprising thing for many is how Compute Savings Plans can automatically cover instances you might not have anticipated. If you suddenly need to spin up a different instance family or in a new region, your CSP commitment can extend to that usage, applying savings without you needing to purchase a new RI. This dynamic application is powered by AWS’s ability to pool your commitment across a broad range of eligible services and instance types.
The next step in optimizing your EC2 spend after understanding Savings Plans and RIs is exploring Spot Instances.