Skip to content

AWS Cost Optimization - SAA-C03

📖 AWS Cost Optimization Pillar - Run systems to deliver business value at the lowest price point

EC2 Cost Optimization

📖 EC2 Pricing - Understand EC2 pricing models

EC2 Pricing Models

On-Demand: - Pay per hour or second - No commitment - Highest cost - Use Case: Short-term, unpredictable workloads

Reserved Instances (RI): - 1 or 3-year commitment - Up to 72% discount vs On-Demand - Standard RI: Highest discount, cannot change instance type - Convertible RI: Change instance family, lower discount - Payment: All upfront, partial upfront, no upfront - Use Case: Steady-state workloads

📖 Reserved Instances - Save up to 72% with long-term commitments

Savings Plans: - 1 or 3-year commitment - Up to 72% discount - Compute Savings Plan: Flexible across instance family, region, OS - EC2 Instance Savings Plan: Flexible within instance family - Use Case: Predictable usage with flexibility

📖 Savings Plans - Flexible pricing model for compute usage

Spot Instances: - Up to 90% discount - Can be terminated with 2-minute warning - Spot Request: One-time or persistent - Spot Fleet: Mix of Spot and On-Demand - Use Case: Fault-tolerant, flexible workloads (batch, data analysis, CI/CD)

Dedicated Hosts: - Physical server dedicated to you - Socket/core visibility for licensing - Most expensive option - Use Case: Regulatory, licensing requirements

Dedicated Instances: - Instances on dedicated hardware - May share hardware with other instances in account - Cheaper than Dedicated Hosts - No access to underlying hardware

Cost Optimization Strategies

Right-Sizing: - Use Compute Optimizer recommendations - Analyze CloudWatch metrics (CPU, memory, network) - Downsize over-provisioned instances - Match instance type to workload

Instance Lifecycle: - Stop instances during off-hours - Terminate unused instances - Use Auto Scaling to match demand - Schedule start/stop with Systems Manager

Purchasing Options Mix: - Reserved/Savings Plans for baseline - On-Demand for variability - Spot for batch and fault-tolerant workloads - Aim for 60-80% Reserved/Savings coverage

Storage Cost Optimization

📖 S3 Storage Classes - Optimize storage costs with the right storage class

S3 Storage Classes

Pricing (per GB, cheapest to most expensive): 1. S3 Glacier Deep Archive: $0.00099/GB (~$1/TB) 2. S3 Glacier Flexible Retrieval: $0.0036/GB (~$3.60/TB) 3. S3 Glacier Instant Retrieval: $0.004/GB (~$4/TB) 4. S3 One Zone-IA: $0.01/GB (~$10/TB) 5. S3 Standard-IA: $0.0125/GB (~$12.50/TB) 6. S3 Intelligent-Tiering: $0.023/GB (~$23/TB) + monitoring 7. S3 Standard: $0.023/GB (~$23/TB)

Lifecycle Policies: - Automatically transition objects between storage classes - Delete old versions or expired objects - Reduce costs for infrequently accessed data

Example Lifecycle: 

0-30 days: S3 Standard
30-90 days: S3 Standard-IA
90-365 days: S3 Glacier Instant Retrieval
365+ days: S3 Glacier Deep Archive

S3 Intelligent-Tiering: - Automatic cost optimization - No retrieval fees for frequent/infrequent tiers - 4 tiers: Frequent, Infrequent (30 days), Archive (90 days), Deep Archive (180 days) - Small monitoring fee per object - Use Case: Unknown or changing access patterns

EBS Cost Optimization

Volume Types: - gp3: Cheaper than gp2, provision IOPS/throughput independently - st1/sc1: Cheaper for throughput-optimized workloads - Delete old snapshots - Resize volumes to actual usage

Strategies: - Use gp3 instead of gp2 (20% cheaper) - Delete unattached volumes - Use snapshot lifecycle policies - Stop instances instead of keeping running with attached EBS

EFS Cost Optimization

  • EFS Standard-IA: 92% cheaper than Standard
  • Lifecycle management: Move to IA after 7, 14, 30, 60, or 90 days
  • One Zone storage classes: 47% savings (if AZ redundancy not required)
  • Elastic throughput for variable workloads

Database Cost Optimization

RDS Strategies

  • Reserved Instances: 1 or 3-year commitment
  • Aurora Serverless v2: Pay per second, auto-scaling
  • Right-size instances: Use CloudWatch metrics
  • Stop dev/test instances during off-hours
  • Read Replicas: Offload read traffic from primary

DynamoDB Strategies

  • On-Demand: Unpredictable traffic (pay per request)
  • Provisioned: Predictable traffic (pre-purchase capacity)
  • Reserved Capacity: 1-year commitment for provisioned
  • Auto Scaling: Match capacity to demand
  • TTL: Automatically delete expired items
  • DAX: Reduce read costs by caching

Redshift Strategies

  • Reserved Nodes: Up to 75% discount
  • Serverless: Pay for usage, no cluster management
  • Pause/Resume: Dev/test clusters
  • Spectrum: Query S3 data directly (avoid loading all data)

Data Transfer Costs

Data Transfer Pricing

  • Inbound: FREE (data coming into AWS)
  • Outbound to Internet: $0.09/GB (varies by region)
  • Between AZs: $0.01-0.02/GB
  • Between Regions: $0.02/GB
  • CloudFront to Internet: $0.085/GB (slightly cheaper + better performance)
  • S3 Transfer Acceleration: Additional $0.04-0.08/GB

Cost Optimization Strategies

CloudFront: - Use CloudFront to reduce data transfer costs - Cache at edge locations - Regional edge caches reduce origin fetches

S3 Transfer Acceleration: - Only when speed matters - Disable if not needed

VPC Endpoints: - FREE for S3 and DynamoDB - Avoid NAT Gateway charges - Keep traffic on AWS network

DirectConnect: - Lower data transfer costs for large, consistent transfers - $0.02/GB vs $0.09/GB for internet transfer

Inter-Region Optimization: - Minimize cross-region data transfer - Use same-region resources when possible - CloudFront for global distribution

Monitoring and Cost Management

📖 AWS Cost Explorer - Visualize and analyze costs

AWS Cost Explorer

  • Visualize spending patterns
  • Filter by service, account, tag, region
  • Forecasting (up to 12 months)
  • Reserved Instance recommendations
  • Savings Plans recommendations

AWS Budgets

  • Set custom cost and usage budgets
  • Alert when exceeding threshold
  • Budget Types:
  • Cost budgets
  • Usage budgets
  • RI utilization budgets
  • RI coverage budgets
  • Savings Plans budgets

📖 AWS Budgets - Set custom budgets and receive alerts

Budget Actions: - Apply IAM policy to restrict services - Stop EC2 or RDS instances - Target specific users, roles, groups

AWS Cost and Usage Report (CUR)

  • Most detailed billing information
  • Hourly, daily, or monthly granularity
  • Delivered to S3
  • Integration with Athena, QuickSight, Redshift
  • Resource-level detail with tags

Cost Allocation Tags

  • Track costs by project, team, environment
  • AWS-generated tags (e.g., aws:createdBy)
  • User-defined tags (e.g., Project:WebApp)
  • Activate tags in Billing Console
  • Apply consistently across resources

AWS Compute Optimizer

  • ML-based recommendations
  • Right-size EC2, Auto Scaling, EBS, Lambda, ECS/Fargate
  • Performance risk assessment
  • Historical utilization analysis
  • Integration with Cost Explorer

📖 AWS Compute Optimizer - Get recommendations to optimize AWS resources

Architectural Cost Optimization

Serverless Architectures

  • Lambda: Pay per request and duration
  • DynamoDB: Pay for read/write capacity
  • S3: Pay for storage and requests
  • No idle capacity costs

Auto Scaling

  • Match capacity to demand
  • Scale down during low usage
  • Prevent over-provisioning
  • Use target tracking policies

Caching

  • ElastiCache: Reduce database load
  • CloudFront: Reduce origin requests
  • DAX: DynamoDB caching
  • Lower compute and data transfer costs

Managed Services

  • RDS vs self-managed database on EC2
  • EKS/ECS vs self-managed Kubernetes
  • Less operational overhead
  • Often cheaper total cost of ownership

Free Tier and Trials

📖 AWS Free Tier - Explore AWS with free tier offerings

AWS Free Tier

Always Free: - Lambda: 1M requests/month - DynamoDB: 25 GB storage, 25 RCU/WCU - CloudWatch: 10 metrics, 10 alarms - SNS: 1M publishes - CloudFront: 1 TB out, 10M requests

12 Months Free: - EC2: 750 hours/month t2.micro or t3.micro - S3: 5 GB Standard storage - RDS: 750 hours/month db.t2.micro - EBS: 30 GB General Purpose (gp2/gp3)

Trials: - Inspector: 90 days - GuardDuty: 30 days - Macie: 30 days

Cost Optimization Best Practices

1. Right-Sizing

  • Continuously analyze and adjust resource sizes
  • Use Compute Optimizer recommendations
  • Monitor CloudWatch metrics
  • Review quarterly

2. Increase Elasticity

  • Use Auto Scaling
  • Serverless architectures
  • Spot instances for fault-tolerant workloads
  • Shut down dev/test environments

3. Choose Right Pricing Model

  • Reserved Instances or Savings Plans for stable workloads
  • Spot for batch and interruptible workloads
  • On-Demand for spiky, unpredictable workloads

4. Optimize Storage

  • Lifecycle policies for S3
  • Delete old snapshots
  • Use EBS gp3 instead of gp2
  • Implement data retention policies

5. Monitor and Analyze

  • Set up billing alerts
  • Review Cost Explorer regularly
  • Tag resources for cost allocation
  • Use AWS Budgets

6. Architecture Efficiency

  • Minimize data transfer
  • Use caching (CloudFront, ElastiCache, DAX)
  • Choose appropriate regions
  • VPC endpoints for S3/DynamoDB

7. Leverage AWS Tools

  • Trusted Advisor cost optimization checks
  • Cost Explorer recommendations
  • Compute Optimizer for rightsizing
  • AWS Cost Anomaly Detection

Exam Tips

Common Cost Optimization Scenarios

  • Steady-state workload: Reserved Instances or Savings Plans
  • Batch processing, fault-tolerant: Spot Instances
  • Spiky, unpredictable: On-Demand + Auto Scaling
  • Infrequent S3 access: S3 IA or Intelligent-Tiering
  • Long-term archive: S3 Glacier Deep Archive
  • Reduce data transfer: CloudFront, VPC Endpoints
  • Right-size resources: Compute Optimizer
  • Unused resources: Delete unattached EBS, old snapshots, idle load balancers

Key Decision Points

  1. Workload predictability → Pricing model
  2. Access pattern → Storage class
  3. Fault tolerance → Spot viability
  4. Data transfer volume → Transfer strategy
  5. Idle periods → Auto Scaling or scheduled stop/start

Cost vs Performance Trade-offs

  • Lower RTO/RPO: Higher cost (multi-site vs backup)
  • Higher availability: Higher cost (multi-AZ)
  • Better performance: Higher cost (Provisioned IOPS, larger instances)
  • Balance based on business requirements

Common Pitfalls

  • Unused Elastic IPs ($0.005/hour)
  • Unattached EBS volumes
  • Old EBS snapshots
  • Data transfer between AZs
  • Over-provisioned resources
  • NAT Gateway charges (use VPC endpoints when possible)
  • Load balancers with no targets