Module 12: Optimizing Cost - Advanced Architecting on AWS

🎯 Module Objectives

Analyze spending with Cost Explorer and cost allocation tags
Set budgets and alerts with AWS Budgets
Right-size instances with Compute Optimizer
Implement Savings Plans and Reserved Instances
Design Spot Instance strategies for fault-tolerant workloads

📊 Cost Management Tools

Tool	Purpose	Key Feature
Cost Explorer	Visualize & analyze spend	Filters by service, account, tag. Forecasting.
AWS Budgets	Alert on thresholds	Cost, usage, RI/SP coverage alerts. Auto-actions.
Cost & Usage Report (CUR)	Granular billing data	Hourly line-item data → S3 → Athena/QuickSight
Compute Optimizer	Right-sizing recommendations	ML-based analysis of CPU/memory utilization
Cost Anomaly Detection	Detect unexpected spend	ML monitors spending patterns, alerts on anomalies

💰 Savings Plans vs. Reserved Instances

Aspect	Savings Plans	Reserved Instances
Commitment	$/hour spend commitment	Specific instance type/region
Flexibility	Compute SP: any instance family/region/OS	Convertible RI: change attributes (lower discount)
Discount	Up to 72% (EC2 Instance SP)	Up to 72% (Standard RI)
Applies to	EC2, Fargate, Lambda, SageMaker	EC2, RDS, OpenSearch, ElastiCache, Redshift
Term	1 or 3 years	1 or 3 years
Recommendation	Start here — simpler, flexible	Use for RDS/ElastiCache (no SP available)

Rule of thumb: Savings Plans for compute (EC2, Fargate, Lambda). RIs for databases (RDS, ElastiCache). Always start with Compute Savings Plan for maximum flexibility.

🎯 Spot Instances

Discount: Up to 90% off On-Demand pricing
Interruption: AWS can reclaim with 2-minute warning
Strategies: Capacity-optimized (fewer interruptions) or lowest-price
Fleet diversification: Use multiple instance types/sizes/AZs to reduce interruption risk
Best for: Batch jobs, CI/CD builds, data processing, stateless web servers, HPC
Not for: Databases, single-instance critical apps, stateful workloads

Pattern: ASG with mixed instances (On-Demand base + Spot for scaling). If Spot is interrupted, ASG replaces with On-Demand. Combine with Spot placement score for optimal AZ selection.

📏 Right-Sizing & Cost Allocation

Compute Optimizer: Analyzes 14 days of CloudWatch metrics → recommends instance type changes
Trusted Advisor: Identifies idle/underutilized resources (EC2, EBS, RDS, ELB)
Cost allocation tags: User-defined tags (Project, Team, Environment) for cost attribution
Organizational cost management: Per-OU budgets, linked account consolidated billing
Auto-scaling: Scale down during off-peak → significant savings for dev/test environments

Quick wins: 1) Delete unused EBS volumes and snapshots. 2) Right-size over-provisioned instances. 3) Stop dev/test environments overnight. 4) Use S3 lifecycle policies. These alone can save 20-40%.

💻 Demo: Cost Optimization

# Get Compute Optimizer recommendations
aws compute-optimizer get-ec2-instance-recommendations \
  --query 'instanceRecommendations[].{Instance:instanceArn,Current:currentInstanceType,Recommended:recommendationOptions[0].instanceType}'

# Check Savings Plan recommendations
aws cost-explorer get-savings-plans-purchase-recommendation \
  --savings-plans-type COMPUTE_SP --term-in-years ONE_YEAR \
  --payment-option NO_UPFRONT --lookback-period-in-days SIXTY_DAYS

# Create a budget with alert
aws budgets create-budget --account-id 123456789012 \
  --budget file://monthly-budget.json \
  --notifications-with-subscribers file://alerts.json

🧪 Knowledge Check

Q1: A company runs various EC2 instance types across multiple regions and wants maximum flexibility in their commitment discount. What should they use?

A) Standard RIs B) Convertible RIs C) Compute Savings Plans D) EC2 Instance Savings Plans

C) Compute Savings Plans — Compute SPs apply across any instance family, size, OS, tenancy, or Region. Maximum flexibility with good discounts (up to 66%). EC2 Instance SPs are locked to a specific family+region.

Q2: Which strategy minimizes Spot interruption risk for a batch processing workload?

A) Use one instance type in one AZ B) Diversify across multiple instance types and AZs with capacity-optimized allocation C) Use only largest instances D) Set a max price

B) Diversify across multiple instance types and AZs with capacity-optimized allocation — Fleet diversification accesses more Spot capacity pools. Capacity-optimized strategy selects pools least likely to be interrupted. Setting max price alone doesn't reduce interruption risk.

📝 Module 12 Summary

Visibility

Cost Explorer, Budgets (alerts + auto-actions), CUR for granular data, Cost Anomaly Detection (ML).

Commitment Discounts

Savings Plans (flexible, compute). RIs (databases). 1 or 3 year terms. Up to 72% off.

Spot & Right-Sizing

Spot up to 90% off (fault-tolerant). Compute Optimizer for right-sizing. Delete unused resources.

Tagging

Cost allocation tags for attribution. Enforce via SCPs. Report by project, team, environment.