AWS Pricing Models: SAA-C03 Study Guide
Understanding AWS pricing models is fundamental to the Cost Optimization pillar of the Well-Architected Framework. For the SAA-C03 exam, you must not only know the names of the models but also identify which model fits a specific business requirement or technical workload.
The “Commuter” Analogy
Think of AWS Pricing like choosing how to get to work:
- On-Demand: Taking a Taxi. You pay exactly for the distance traveled. No commitment, but the highest cost per mile.
- Reserved Instances/Savings Plans: Monthly Bus Pass. You commit to using the service frequently for 1–3 years in exchange for a massive discount.
- Spot Instances: Standby Tickets. You get a huge discount on a flight, but if a full-fare passenger shows up, you might lose your seat with very short notice.
Core Concepts & Well-Architected Lens
AWS pricing is built around three main drivers: Compute, Storage, and Data Out. Inbound data transfer and data transfer between services in the same AZ are typically free, while “Data Transfer Out” to the internet is a primary cost driver.
The Cost Optimization Pillar
The exam focuses on “Right Sizing” and “Increasing Elasticity.” AWS expects architects to select the lowest-cost resources that still meet performance requirements. This involves choosing the correct pricing model based on predictability and fault tolerance.
Comparison of Compute Pricing Models
| Model | Commitment | Discount | Best For… |
|---|---|---|---|
| On-Demand | None (Per second/hour) | 0% (Baseline) | Short-term, unpredictable workloads, or dev/test. |
| Reserved Instances (RI) | 1 or 3 Years | Up to 72% | Steady-state, predictable usage (e.g., Production DBs). |
| Savings Plans | 1 or 3 Years ($/hr) | Up to 72% | Flexibility across EC2, Fargate, and Lambda. |
| Spot Instances | None (Can be reclaimed) | Up to 90% | Fault-tolerant, stateless, or batch processing. |
Scenario-Based Learning: Decision Matrix
IF the requirement is a 24/7 Production Database THEN use Reserved Instances.
IF the requirement is a Big Data batch job that can restart THEN use Spot Instances.
IF the workload is migrating frequently between Lambda and Fargate THEN use Compute Savings Plans.
IF you are testing a new application for 2 weeks THEN use On-Demand.
Exam Tips: Golden Nuggets
- Spot for Stateless: If the exam mentions “Cost-Effective” and “Stateless” or “Fault-Tolerant,” the answer is almost always Spot Instances.
- Convertible RIs: Use these if you need the discount but might need to change the Instance Family (e.g., moving from T3 to M5) later.
- The “Data Out” Trap: Always remember that Data Transfer Into AWS is free, but Data Transfer Out to the internet is where costs scale.
- Standard vs. Compute Savings Plans: EC2 Instance Savings Plans offer higher discounts but are locked to a specific family; Compute Savings Plans are more flexible.
Visual Guide: AWS Cost Architecture
Strategic Selection for SAA-C03 Success
Key Services
EC2: Multiple models available.
S3: Tiered pricing (Standard, IA, Glacier).
Lambda: Pay-per-request and duration.
Common Pitfalls
Over-provisioning: Paying for idle capacity.
Ignoring Data Out: Budgeting for compute but forgetting transfer fees.
Quick Patterns
CI/CD: Use Spot for build agents.
Production: Use 3-year RI for core DB instances.
Startups: Use Savings Plans for flexibility.