AWS Cost Optimization: 6 Strategies That Actually Work

AWS bills can spiral out of control fast. After helping organizations cut their cloud spend by 30-50%, these are the strategies that consistently deliver results.

Reserved Instances for Predictable Workloads

Reserved Instances (RIs) offer up to 75% savings on EC2 and RDS compared to on-demand pricing. The catch: you're committing to one or three years.

When to use RIs:

• Production databases that run 24/7
• Baseline application servers with consistent load
• Long-running batch processing systems

Implementation approach:

1. Pull 3-6 months of usage data from Cost Explorer
2. Identify instances running at 70%+ utilization consistently
3. Start with Standard RIs for stable workloads
4. Use Convertible RIs when you anticipate architecture changes

Monitor RI utilization monthly. Unused reservations are wasted money.

Cost Explorer: Your First Line of Defense

Cost Explorer reveals spending patterns that aren't obvious from monthly bills. Set it up properly:

Essential reports to create:

• Daily costs by service (catch anomalies early)
• Costs by linked account (identify runaway accounts)
• Costs by tag (attribute spending to teams/projects)

Pro tip: Enable cost anomaly detection. AWS will alert you when spending deviates from historical patterns - often catching issues before they become expensive.

Auto Scaling Done Right

Auto Scaling prevents overprovisioning, but misconfigured scaling policies waste money or cause outages.

Scaling policy checklist:

• Set minimum instances based on your baseline traffic
• Use target tracking scaling with 60-70% CPU target
• Configure scale-in cooldown to prevent thrashing
• Test your scaling policies before production deployment

# Example target tracking policy
TargetTrackingScalingPolicy:
  TargetValue: 65.0
  PredefinedMetricType: ASGAverageCPUUtilization
  ScaleInCooldown: 300
  ScaleOutCooldown: 60

Savings Plans vs Reserved Instances

Savings Plans offer more flexibility than RIs with similar discounts (up to 72%).

Choose Compute Savings Plans when:

• Running diverse workloads across regions
• Using both EC2 and Fargate
• Planning infrastructure changes

Choose EC2 Instance Savings Plans when:

• Workloads are stable and well-understood
• Maximum discount is priority
• Instance families won't change

Calculate your commitment carefully. Over-committing locks you into paying for capacity you don't use.

CloudWatch Alarms for Cost Control

Set billing alarms before costs become problems:

aws cloudwatch put-metric-alarm \
  --alarm-name "Monthly-Bill-Threshold" \
  --metric-name EstimatedCharges \
  --namespace AWS/Billing \
  --statistic Maximum \
  --period 21600 \
  --threshold 1000 \
  --comparison-operator GreaterThanThreshold \
  --alarm-actions arn:aws:sns:us-east-1:ACCOUNT:billing-alerts

Beyond billing, monitor resource utilization:

• EC2 instances under 10% CPU for extended periods
• RDS instances with minimal connections
• EBS volumes with zero IOPS

Automated Resource Cleanup

Unused resources accumulate fast. Automate their removal:

High-impact targets:

• Unattached EBS volumes (charged even when not mounted)
• Old EBS snapshots (often forgotten after AMI creation)
• Idle Elastic IPs (charged when not associated)
• Unused NAT Gateways ($32/month each)

Lambda function for cleanup:

import boto3

def cleanup_unattached_volumes(event, context):
    ec2 = boto3.client('ec2')
    volumes = ec2.describe_volumes(
        Filters=[{'Name': 'status', 'Values': ['available']}]
    )

    for vol in volumes['Volumes']:
        # Check if volume has been unattached for 7+ days
        # Add your deletion logic with appropriate safeguards
        pass

Schedule this with EventBridge for weekly execution.

Key Takeaways

• Start with visibility - You can't optimize what you can't measure. Cost Explorer is free.
• Reserved Instances/Savings Plans - Easy wins for predictable workloads. Start with 1-year terms.
• Auto Scaling - Right-size dynamically, but test your policies thoroughly.
• Automate cleanup - Manual audits don't scale. Build automated resource hygiene.
• Monitor continuously - Set billing alerts before you need them.

Cost optimization isn't a one-time project. Build it into your operational rhythm with monthly reviews and automated guardrails.