적정 규모 조정 및 Compute Optimizer

과다 프로비저닝 문제

# Common over-provisioning symptoms:
# EC2: average CPU < 10%, memory < 20%
# RDS: storage auto-grow never triggered
# Lambda: allocated memory rarely exceeds 40%
# EBS: provisioned IOPS consistently unused

# Cost impact example:
# Over-provisioned r5.8xlarge at $1.92/hr: $1,382/month
# Correct size r5.xlarge at $0.252/hr:      $181/month
# Savings: $1,201/month per instance

AWS Compute Optimizer 개요

# Opt in to Compute Optimizer
aws compute-optimizer update-enrollment-status \
  --status Active

# Get EC2 recommendations
aws compute-optimizer get-ec2-instance-recommendations \
  --query 'instanceRecommendations[].{Instance:currentInstanceType,Recommended:recommendationOptions[0].instanceType,Savings:recommendationOptions[0].estimatedMonthlySavings.value,Risk:recommendationOptions[0].performanceRisk}'

# Get Lambda recommendations
aws compute-optimizer get-lambda-function-recommendations

Compute Optimizer를 사용한 EC2 적정 규모 설정

# EC2 recommendation findings:
# Finding: OVER_PROVISIONED
# CurrentInstanceType: m5.4xlarge
# RecommendedInstanceType: m5.xlarge
# CPU utilisation: P99 = 22%, P50 = 4%
# Memory utilisation: P99 = 18%, P50 = 8%
# EstimatedMonthlySavings: $380
# PerformanceRisk: VeryLow

# Always check:
# - Does current instance have burstable credits? (T3)
# - Are there workload spikes not captured in averages?
# - Is there seasonality to consider?

Lambda 적정 규모 설정

# Compute Optimizer Lambda recommendations
aws compute-optimizer get-lambda-function-recommendations \
  --function-arns arn:aws:lambda:us-east-1:123:function:my-function

# Response shows:
# currentMemorySize: 1024 MB
# recommendedMemorySize: 256 MB
# utilizationMetrics:
#   - type: MEMORY_MAXIMUM, value: 89 (MB)
# estimatedMonthlySavings: $45

# 89 MB actual vs 1024 MB allocated = 935 MB wasted

EBS 볼륨 적정 규모 설정

# Get EBS volume recommendations
aws compute-optimizer get-ebs-volume-recommendations \
  --volume-arns arn:aws:ec2:us-east-1:123:volume/vol-12345

# Common finding: gp2 -> gp3 migration
# gp2: 500 GB = 500 GB * $0.10 = $50/month
#   IOPS: 1,500 (3 IOPS/GB, not configurable)
# gp3: 500 GB = $40/month + 3,000 IOPS free
#   Save $10/month plus get MORE baseline IOPS

# Find unattached EBS volumes
aws ec2 describe-volumes \
  --filters Name=status,Values=available \
  --query 'Volumes[].{VolumeId:VolumeId,Size:Size,Created:CreateTime}'

autoScalingGroup 추천

# Get ASG recommendations
aws compute-optimizer get-auto-scaling-group-recommendations \
  --auto-scaling-group-arns arn:aws:autoscaling:us-east-1:123:autoScalingGroup:abc:autoScalingGroupName/my-asg

# If recommendation is to switch to Graviton:
# Current: m5.large (x86, $0.096/hr)
# Recommended: m6g.large (Graviton2, $0.077/hr)
# Savings: 20% per instance
# ASG at 10 instances average = $220/month savings

비용 분석을 위한 Trusted Advisor

# Trusted Advisor: get cost optimisation checks
aws support describe-trusted-advisor-checks \
  --language en \
  --query 'checks[?category==`cost_optimizing`].{Name:name,Id:id}'

# Key checks:
# Qkj5MU5fNp: Low utilisation EC2 instances
# Z4AUBRNSmh: Unassociated Elastic IP addresses
# DAvU99Dc4C: Underutilized EBS volumes
# hjLMh88uM8: Idle load balancers
# 1e93e4c0b5: Unused reserved instances

인스턴스 세대 업그레이드

# EC2 instance generation comparison (same price tier):
# m5.large:  2 vCPU, 8 GB, $0.096/hr  (2019)
# m6i.large: 2 vCPU, 8 GB, $0.096/hr  (2021, ~10% faster)
# m7i.large: 2 vCPU, 8 GB, $0.1008/hr (2023, ~15% faster)
# m7g.large: 2 vCPU, 8 GB, $0.0808/hr (2023, Graviton3, cheapest)

# Upgrade path for Linux workloads:
# m5 -> m7g (Graviton): best price/performance
# m5 -> m7i: same architecture, no code changes

RDS 인스턴스 적정 규모 설정

# Monitor RDS utilisation for right-sizing
aws cloudwatch get-metric-statistics \
  --namespace AWS/RDS \
  --metric-name CPUUtilization \
  --dimensions Name=DBInstanceIdentifier,Value=mydb \
  --start-time 2026-05-21T00:00:00Z \
  --end-time 2026-06-21T00:00:00Z \
  --period 86400 \
  --statistics Maximum Average

# If 30-day P99 CPU < 30%, consider downsizing
# If P99 CPU > 80%, consider upsizing

적정 규모 설정의 운영화

# Script to release unassociated Elastic IPs
aws ec2 describe-addresses \
  --query 'Addresses[?!AssociationId].AllocationId' \
  --output text | xargs -I {} \
  aws ec2 release-address --allocation-id {}

# Script to delete unattached EBS volumes (careful!)
# First check if snapshots exist before deleting
aws ec2 describe-volumes \
  --filters Name=status,Values=available \
  --query 'Volumes[].VolumeId' \
  --output text

적정 규모 설정과 아키텍처 변경 비교

# Cost optimisation hierarchy:
# Level 1: Right-sizing (quick wins, days)
#   - Compute Optimizer recommendations
#   - Delete unused resources
#   - Elastic IP, EBS cleanup

# Level 2: Purchasing model (weeks)
#   - Reserved Instances / Savings Plans
#   - Spot for eligible workloads

# Level 3: Architecture (months)
#   - Serverless migration
#   - Container consolidation
#   - Caching layer addition
#   - Database optimisation

빠른 확인

단원 요약