AWS Certified Solutions Architect Associate (SAA-C03) Domain 4

Design Cost-Optimized Architectures

Official Exam Guide: SAA-C03 Exam Guide

---

Domain Overview

Domain Weight: 20% of the exam

This domain tests your ability to design cost-effective storage, compute, database, and networking solutions.

---

Task 4.1: Design cost-optimized storage solutions

S3 Storage Optimization

S3 Storage Classes

Selection by Access Pattern:

Access Pattern	Storage Class	Use Case
Frequent	S3 Standard	Active data
Unknown/changing	S3 Intelligent-Tiering	Unpredictable access
Infrequent	S3 Standard-IA	Backups, disaster recovery
Infrequent, single AZ	S3 One Zone-IA	Reproducible data
Archive	S3 Glacier Instant Retrieval	Quarterly access, milliseconds
Archive	S3 Glacier Flexible Retrieval	Minutes to hours retrieval
Long-term archive	S3 Glacier Deep Archive	Once/twice per year, 12 hours

Cost Optimization:

• Use Intelligent-Tiering for unknown patterns
• Lifecycle policies to transition objects
• Delete incomplete multipart uploads
• Use S3 Storage Lens for insights
• Request Payer for user-pays scenarios

AWS Documentation:

• S3 Storage Classes

S3 Lifecycle Policies

Example Policy:

Day 0: S3 Standard
Day 30: S3 Standard-IA
Day 90: S3 Glacier Flexible Retrieval
Day 365: S3 Glacier Deep Archive
Day 2555: Delete

Best Practices:

• Transition to IA after 30 days minimum
• Use filters (prefix, tags)
• Test policies before applying
• Monitor cost savings

EBS Optimization

Volume Selection:

• gp3 - Most cost-effective general purpose (vs gp2)
• st1 - Cost-effective throughput (vs io2 for sequential)
• sc1 - Lowest cost for infrequent access

Cost Reduction:

• Delete unattached volumes
• Snapshot old volumes, delete volume
• Delete old snapshots (use lifecycle)
• Use gp3 instead of gp2 (same price, better performance)
• Right-size volumes

AWS Documentation:

• EBS Pricing

EFS Optimization

Cost Reduction:

• Enable Lifecycle Management (IA transition)
• Use One Zone storage class when appropriate
• Use Bursting instead of Provisioned throughput
• Delete unused file systems

AWS Documentation:

• EFS Pricing

---

Task 4.2: Design cost-optimized compute solutions

EC2 Purchase Options

Pricing Models Comparison

Model	Cost Savings	Commitment	Best For
On-Demand	Baseline (1x)	None	Variable, short-term
Reserved (1yr)	~30-40%	1 year	Steady state
Reserved (3yr)	~50-60%	3 years	Long-term predictable
Savings Plans	~30-60%	1-3 years	Flexible compute
Spot	~50-90%	None	Fault-tolerant workloads
Dedicated Hosts	Varies	Varies	Compliance, licensing

Reserved Instances (RIs)

Types:

• Standard RI - Highest savings, can’t change instance type
• Convertible RI - Can change instance type, lower savings

Payment Options:

• All Upfront (highest savings)
• Partial Upfront
• No Upfront (lowest savings)

Best Practices:

• Analyze usage with Cost Explorer
• Buy for steady-state workloads
• Start with 1-year, then 3-year
• Use RI marketplace to sell unused

Savings Plans

Types:

• Compute Savings Plans - Most flexible (EC2, Fargate, Lambda)
• EC2 Instance Savings Plans - EC2 only, specific instance family

Benefits:

• More flexible than RIs
• Apply automatically
• Combine with Spot for additional savings

Spot Instances

Key Concepts:

• Up to 90% discount
• Can be interrupted with 2-minute warning
• Spot Fleet manages multiple Spot Instances
• Spot Blocks for defined duration (deprecated)

Use Cases:

• Batch processing
• Big data analysis
• CI/CD pipelines
• Containerized workloads
• Fault-tolerant applications

Best Practices:

• Use multiple instance types and AZs
• Implement graceful shutdown
• Use Spot Fleet for resilience
• Combine with On-Demand for base capacity

AWS Documentation:

• EC2 Pricing

Instance Right-Sizing

Optimization Steps:

1. Enable CloudWatch detailed monitoring
2. Analyze CPU, memory, network utilization
3. Use AWS Compute Optimizer
4. Downsize underutilized instances
5. Use Auto Scaling to match demand

Tools:

• AWS Compute Optimizer
• AWS Cost Explorer
• Trusted Advisor

Lambda Cost Optimization

Cost Factors:

• Number of requests
• Duration (GB-seconds)
• Memory allocation

Optimization:

• Optimize function memory (affects CPU)
• Reduce package size
• Use ARM (Graviton2) for lower cost
• Reduce execution time
• Use Provisioned Concurrency carefully

AWS Documentation:

• Lambda Pricing

Container Optimization

ECS/EKS Cost Reduction:

• Use Fargate Spot for cost savings
• Right-size task definitions
• Use Compute Savings Plans
• Enable Cluster Auto Scaling

AWS Documentation:

• Fargate Pricing

---

Task 4.3: Design cost-optimized database solutions

RDS Cost Optimization

Purchase Options:

• On-Demand - Pay per hour
• Reserved Instances - Up to 69% savings

Cost Reduction:

• Use appropriate instance size
• Use read replicas instead of Multi-AZ for read scaling
• Stop instances when not needed (dev/test)
• Use Aurora Serverless for variable workloads
• Delete automated backups for terminated instances
• Use Aurora vs RDS for better performance/cost

Aurora Serverless:

• Auto-scales based on load
• Pay per ACU (Aurora Capacity Unit)
• Pause when inactive
• Use case: Infrequent, unpredictable workloads

AWS Documentation:

• RDS Pricing

DynamoDB Cost Optimization

Capacity Modes:

• On-Demand - Pay per request (unpredictable)
• Provisioned - Reserve capacity (predictable, lower cost)

Cost Reduction:

• Use on-demand for new/unpredictable workloads
• Switch to provisioned for predictable patterns
• Enable auto-scaling for provisioned
• Use Global Tables only when needed
• Delete unused tables and indexes
• Use DynamoDB Standard-IA for infrequent access

AWS Documentation:

• DynamoDB Pricing

ElastiCache Cost Optimization

Cost Reduction:

• Use appropriate node size
• Use Reserved Nodes for steady workloads
• Scale horizontally with read replicas
• Delete unused clusters

AWS Documentation:

• ElastiCache Pricing

Redshift Cost Optimization

Cost Reduction:

• Use RA3 nodes with managed storage
• Pause and resume clusters (dev/test)
• Use Reserved Nodes
• Concurrency Scaling (pay for peak usage only)
• Use Spectrum for S3 data (avoid loading)

AWS Documentation:

• Redshift Pricing

---

Task 4.4: Design cost-optimized network architectures

Data Transfer Costs

Cost Factors:

• Data IN to AWS: Free
• Data OUT to internet: Charged
• Data between regions: Charged
• Data within same AZ: Free (usually)
• Data between AZs: Charged

Optimization:

• Use CloudFront to reduce data transfer
• Keep data in same region when possible
• Use VPC endpoints (no NAT Gateway fees)
• Use Direct Connect for large transfers
• Compress data before transfer

VPC Endpoints

Types:

• Gateway Endpoints (S3, DynamoDB) - Free
• Interface Endpoints (other services) - Hourly + data charge

Cost Savings:

• No NAT Gateway fees for S3/DynamoDB
• No data transfer charges to internet
• More secure

AWS Documentation:

• VPC Endpoints

NAT Gateway vs NAT Instance

NAT Gateway:

• Managed, highly available
• ~$30-45/month + data processing
• Recommended for production

NAT Instance:

• Self-managed EC2 instance
• Lower cost for small workloads
• Less reliable, requires maintenance

Cost Optimization:

• Use VPC endpoints for S3/DynamoDB (avoid NAT)
• Consider NAT instance for dev/test
• One NAT Gateway per AZ (not per subnet)

CloudFront Cost Optimization

Cost Factors:

• Data transfer OUT
• HTTP/HTTPS requests
• Price classes (geographic regions)

Optimization:

• Use lower price class if users are regional
• Enable compression
• Cache aggressively
• Use Origin Shield for multiple origins

AWS Documentation:

• CloudFront Pricing

---

General Cost Optimization Strategies

Monitoring and Analysis

Tools:

• AWS Cost Explorer - Visualize and analyze costs
• AWS Budgets - Set custom budgets and alerts
• AWS Cost Anomaly Detection - ML-based anomaly detection
• AWS Trusted Advisor - Cost optimization recommendations
• AWS Compute Optimizer - Right-sizing recommendations

Best Practices:

• Tag all resources (cost allocation tags)
• Set up billing alerts
• Review Cost Explorer monthly
• Use Budgets for forecasting
• Enable Cost Anomaly Detection

Architectural Best Practices

Design Principles:

• Right-sizing from the start
• Increase elasticity (Auto Scaling)
• Choose right pricing models
• Optimize over time
• Use managed services when cost-effective

Common Optimizations:

• Auto Scaling to match demand
• S3 Lifecycle policies
• Delete unused resources
• Use Spot for fault-tolerant workloads
• Reserved capacity for steady state
• CloudFront for content delivery
• Serverless for variable workloads

---

Exam Tips

Common Question Patterns:

1. Reduce storage costs → S3 Lifecycle policies, Intelligent-Tiering
2. Cost-effective compute for steady workload → Reserved Instances or Savings Plans
3. Fault-tolerant batch processing → Spot Instances
4. Variable workload database → Aurora Serverless or DynamoDB On-Demand
5. Reduce data transfer costs → CloudFront, VPC endpoints
6. Dev/test environment → Start/stop resources, Spot Instances
7. Long-term archive → S3 Glacier Deep Archive

Cost Optimization Framework:

1. Measure - Use Cost Explorer, tags
2. Analyze - Identify waste, underutilized resources
3. Optimize - Right-size, use appropriate pricing
4. Monitor - Set budgets, alerts, review regularly

Key Principles:

• Pay only for what you use
• Reduce waste (delete unused resources)
• Use the right resource for the job
• Use Reserved capacity for steady state
• Use Spot for fault-tolerant workloads
• Implement Auto Scaling
• Monitor and optimize continuously

---

Final Thoughts

Domain 4 (Design Cost-Optimized Architectures) is 20% of the exam.

Master these concepts:

1. EC2 pricing models (On-Demand, Reserved, Savings Plans, Spot)
2. S3 storage classes and lifecycle policies
3. RDS Reserved Instances and Aurora Serverless
4. Data transfer costs and VPC endpoints
5. CloudFront for reducing data transfer
6. Cost monitoring tools (Cost Explorer, Budgets)
7. Right-sizing and Auto Scaling

Remember: Cost optimization is continuous, not one-time!