aws-cost-optimizer（AWS cost 优化器）

AWS Cost Optimizer

Analyze and optimize AWS costs with actionable recommendations.

Quick Start

Review Cost Explorer, identify top costs, right-size instances, use Reserved Instances, delete unused resources.

Instructions

Cost Analysis Process

1. Identify cost drivers
2. Analyze usage patterns
3. Recommend optimizations
4. Estimate savings
5. Implement changes

Using AWS Cost Explorer

Access Cost Explorer:

• AWS Console → Billing → Cost Explorer
• View costs by service, region, tag

Key metrics to check:

• Monthly costs by service
• Cost trends over time
• Unused or underutilized resources
• Data transfer costs

EC2 Cost Optimization

Right-sizing instances:

Check utilization:

# Get CloudWatch metrics
aws cloudwatch get-metric-statistics \
  --namespace AWS/EC2 \
  --metric-name CPUUtilization \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 \
  --start-time 2024-01-01T00:00:00Z \
  --end-time 2024-01-31T23:59:59Z \
  --period 3600 \
  --statistics Average

Recommendations:

• CPU < 20%: Downsize instance type
• CPU > 80%: Upsize or add instances
• Memory < 50%: Consider smaller instance

Reserved Instances:

• 1-year: ~30-40% savings
• 3-year: ~50-70% savings
• Best for predictable workloads

Savings Plans:

• More flexible than Reserved Instances
• Commit to $/hour usage
• Apply across instance families

Spot Instances:

• Up to 90% savings
• For fault-tolerant workloads
• Batch processing, CI/CD, testing

Stop unused instances:

# Find stopped instances
aws ec2 describe-instances \
  --filters "Name=instance-state-name,Values=stopped" \
  --query 'Reservations[].Instances[].[InstanceId,Tags[?Key==`Name`].Value|[0]]'

# Terminate if not needed
aws ec2 terminate-instances --instance-ids i-1234567890abcdef0

S3 Cost Optimization

Lifecycle policies:

{
  "Rules": [{
    "Id": "Archive old data",
    "Status": "Enabled",
    "Transitions": [
      {
        "Days": 30,
        "StorageClass": "STANDARD_IA"
      },
      {
        "Days": 90,
        "StorageClass": "GLACIER"
      }
    ],
    "Expiration": {
      "Days": 365
    }
  }]
}

Storage classes:

• Standard: Frequent access
• Standard-IA: Infrequent access (30+ days)
• Glacier: Archive (90+ days)
• Glacier Deep Archive: Long-term archive

Delete incomplete multipart uploads:

aws s3api list-multipart-uploads --bucket my-bucket
# Set lifecycle rule to abort after 7 days

Analyze storage:

# Get bucket size
aws s3 ls s3://my-bucket --recursive --summarize

RDS Cost Optimization

Right-size databases:

• Check CPU, memory, IOPS utilization
• Downsize if consistently low
• Use Aurora Serverless for variable workloads

Reserved Instances:

• 1-year: ~30-40% savings
• 3-year: ~50-60% savings

Stop dev/test databases:

# Stop RDS instance
aws rds stop-db-instance --db-instance-identifier mydb

# Start when needed
aws rds start-db-instance --db-instance-identifier mydb

Delete old snapshots:

# List snapshots
aws rds describe-db-snapshots --query 'DBSnapshots[?SnapshotCreateTime<`2023-01-01`]'

# Delete old snapshots
aws rds delete-db-snapshot --db-snapshot-identifier snapshot-id

Data Transfer Costs

Reduce data transfer:

• Use CloudFront for content delivery
• Keep data in same region
• Use VPC endpoints for AWS services
• Compress data before transfer

VPC endpoints:

# Create S3 VPC endpoint (no data transfer charges)
aws ec2 create-vpc-endpoint \
  --vpc-id vpc-12345678 \
  --service-name com.amazonaws.us-east-1.s3 \
  --route-table-ids rtb-12345678

EBS Cost Optimization

Delete unattached volumes:

# Find unattached volumes
aws ec2 describe-volumes \
  --filters "Name=status,Values=available" \
  --query 'Volumes[].[VolumeId,Size,VolumeType]'

# Delete if not needed
aws ec2 delete-volume --volume-id vol-1234567890abcdef0

Delete old snapshots:

# List old snapshots
aws ec2 describe-snapshots --owner-ids self \
  --query 'Snapshots[?StartTime<`2023-01-01`]'

# Delete
aws ec2 delete-snapshot --snapshot-id snap-1234567890abcdef0

Use gp3 instead of gp2:

• gp3 is 20% cheaper
• Better performance
• Migrate existing volumes

Lambda Cost Optimization

Optimize memory allocation:

• More memory = faster execution = lower cost
• Test different memory settings
• Use AWS Lambda Power Tuning tool

Reduce cold starts:

• Use provisioned concurrency (if needed)
• Keep functions warm with scheduled events
• Minimize dependencies

Monitor invocations:

# Get Lambda metrics
aws cloudwatch get-metric-statistics \
  --namespace AWS/Lambda \
  --metric-name Invocations \
  --dimensions Name=FunctionName,Value=my-function \
  --start-time 2024-01-01T00:00:00Z \
  --end-time 2024-01-31T23:59:59Z \
  --period 86400 \
  --statistics Sum

CloudWatch Costs

Reduce log retention:

# Set log retention to 7 days
aws logs put-retention-policy \
  --log-group-name /aws/lambda/my-function \
  --retention-in-days 7

Delete unused log groups:

# List log groups
aws logs describe-log-groups

# Delete
aws logs delete-log-group --log-group-name /aws/lambda/old-function

Unused Resources

Find unused resources:

Elastic IPs not attached:

aws ec2 describe-addresses \
  --query 'Addresses[?AssociationId==null]'

Load balancers with no targets:

aws elbv2 describe-load-balancers
aws elbv2 describe-target-health --target-group-arn arn

NAT Gateways with low traffic:

# Check CloudWatch metrics for BytesOutToDestination

Cost Optimization Checklist

Compute:

• Right-size EC2 instances
• Use Reserved Instances for predictable workloads
• Use Spot Instances for fault-tolerant workloads
• Stop/terminate unused instances
• Implement auto-scaling

Storage:

• Implement S3 lifecycle policies
• Delete old EBS snapshots
• Delete unattached EBS volumes
• Use appropriate S3 storage classes
• Delete incomplete multipart uploads

Database:

• Right-size RDS instances
• Use Reserved Instances
• Stop dev/test databases when not in use
• Delete old RDS snapshots
• Consider Aurora Serverless

Networking:

• Use CloudFront to reduce data transfer
• Delete unused Elastic IPs
• Use VPC endpoints
• Review NAT Gateway usage

Monitoring:

• Reduce CloudWatch log retention
• Delete unused log groups
• Review custom metrics

General:

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

Savings Estimation

Reserved Instances:

• 1-year, no upfront: ~30% savings
• 1-year, all upfront: ~40% savings
• 3-year, all upfront: ~60% savings

Spot Instances:

• 70-90% savings vs On-Demand

S3 Lifecycle:

• Standard-IA: ~50% cheaper than Standard
• Glacier: ~80% cheaper than Standard

Right-sizing:

• Typical savings: 20-40% on oversized instances

Tools and Commands

AWS Cost Explorer:

• View costs by service, region, tag
• Forecast future costs
• Identify cost anomalies

AWS Budgets:

# Create budget
aws budgets create-budget \
  --account-id 123456789012 \
  --budget file://budget.json

AWS Trusted Advisor:

• Cost optimization recommendations
• Underutilized resources
• Idle resources

Third-party tools:

• CloudHealth
• CloudCheckr
• Spot.io

Best Practices

Tagging strategy:

• Tag all resources
• Use tags for cost allocation
• Common tags: Environment, Project, Owner, CostCenter

Regular reviews:

• Weekly: Check for anomalies
• Monthly: Review Cost Explorer
• Quarterly: Optimize Reserved Instances

Automation:

• Auto-stop dev instances at night
• Auto-delete old snapshots
• Auto-scale based on demand

Monitoring:

• Set up billing alerts
• Use AWS Budgets
• Enable Cost Anomaly Detection