AWS Cost Optimization: Savings Plans vs. Reserved Instances
In the AWS Certified Solutions Architect – Associate (SAA-C03) exam, understanding how to reduce costs for predictable workloads is vital. Both Savings Plans (SP) and Reserved Instances (RI) offer significant discounts (up to 72%) in exchange for a commitment to a consistent amount of usage for a 1 or 3-year term.
The Gym Membership Analogy
Imagine you want to get fit:
- Reserved Instance (RI): This is like renting a specific locker and a specific treadmill (e.g., Treadmill #4) at a specific gym branch. It’s yours whenever you want it, but if you want to use the rowing machine instead, or go to a different branch, your “reservation” doesn’t help you much.
- Savings Plan (SP): This is like a prepaid membership credit. You commit to spending $10 a day at the gym. You can use the treadmill, the pool, or the weights, and you can even go to different gym locations. As long as you spend that $10, you get the discounted rate. It’s much more flexible.
Core Concepts & The Well-Architected Framework
Under the Cost Optimization Pillar of the AWS Well-Architected Framework, AWS encourages “Expenditure Awareness.” RIs and SPs allow architects to move from unblended On-Demand pricing to a committed model once a baseline of usage is established.
What are they?
They are not physical “instances” but rather billing discounts applied to your usage. You commit to a 1 or 3-year term, and in return, AWS lowers the hourly rate.
Why use them?
- Predictability: Lowering the “Effective Cost” of long-running workloads.
- Financial Planning: Choosing between “All Upfront” (highest discount), “Partial Upfront,” or “No Upfront” to manage cash flow.
Detailed Comparison Table
| Feature | Standard RI | Convertible RI | Compute Savings Plan | EC2 Instance Savings Plan |
|---|---|---|---|---|
| Flexibility | Low (Locked to Instance Family) | Medium (Can change family) | Highest (Instance, Region, OS, Fargate) | High (OS, Tenancy within family) |
| Services | EC2, RDS, ElastiCache, Redshift | EC2 Only | EC2, Fargate, Lambda | EC2 Only |
| Regional vs Zonal | Both available | Both available | Regional only | Regional only |
| Capacity Reservation | Yes (If Zonal) | Yes (If Zonal) | No | No |
Scenario-Based Learning: Decision Matrix
- IF you use Fargate or Lambda and want a discount: THEN use Compute Savings Plans.
- IF you need a discount for RDS or Redshift: THEN use Reserved Instances (SPs don’t cover these yet).
- IF you need to guarantee capacity in a specific Availability Zone for a product launch: THEN use Zonal Reserved Instances.
- IF your EC2 instance family changes frequently (e.g., moving from C5 to M5): THEN use Compute Savings Plans.
Exam Tips: Golden Nuggets
- The “Fargate” Trap: If an exam question mentions Fargate cost optimization, “Reserved Instances” is always the wrong answer. Only Savings Plans apply to Fargate.
- Modification vs. Exchange: Standard RIs can be modified (change AZ within region), but only Convertible RIs can be exchanged for different instance families.
- Regional RI vs. Zonal RI: Regional RIs provide a discount but NO capacity reservation. Zonal RIs provide a capacity reservation but less flexibility.
- Queuing: You can “Queue” a Savings Plan to start on a specific date in the future (handy for replacing expiring RIs).
Visualizing Cost Commitment
AWS Savings Models at a Glance
Architectural Flow: Commitment to Discount
Key Services
Savings Plans: EC2, AWS Lambda, AWS Fargate, SageMaker.
Reserved Instances: EC2, RDS, Redshift, ElastiCache, OpenSearch.
Common Pitfalls
- Buying 3-year terms for short-term projects.
- Over-committing $/hour (you pay even if usage drops).
- Assuming RIs apply to Lambda (they don’t).
Quick Patterns
- Dynamic: Use Compute SP for auto-scaling groups.
- Static: Use Standard RI for a DB instance that never changes.
- Global: SPs are easier for multi-region teams.