고성능 및 비용 최적화 시나리오

시나리오 1: 데이터베이스 부하를 줄이기 위한 캐싱

import boto3, json

elasticache = boto3.client('elasticache')
redis_client = None  # assume redis-py client connected to ElastiCache endpoint

def get_article(article_id):
    cache_key = 'article:' + str(article_id)
    # Check cache first
    cached = redis_client.get(cache_key)
    if cached:
        return json.loads(cached)  # cache hit: <1ms
    # Cache miss: query RDS
    article = rds_query('SELECT * FROM articles WHERE id = %s', article_id)
    # Write to cache with 1-hour TTL
    redis_client.setex(cache_key, 3600, json.dumps(article))
    return article

시나리오 2: 정적 자산 전송을 위한 CloudFront

# CloudFront origin for ALB + separate behaviour for static assets
aws cloudfront create-distribution --distribution-config '{
  'Origins': {
    'Quantity': 1,
    'Items': [{
      'Id': 'alb-origin',
      'DomainName': 'my-alb.us-east-1.elb.amazonaws.com',
      'CustomOriginConfig': {"HTTPSPort": 443, "OriginProtocolPolicy": "https-only"}
    }]
  },
  'CacheBehaviors': {
    'Quantity': 1,
    'Items': [{
      'PathPattern': '/static/*',
      'DefaultTTL': 604800,
      'MaxTTL': 604800
    }]
  },
  'DefaultCacheBehavior': {"DefaultTTL": 0}
}'

시나리오 3: Compute Optimizer를 사용한 적정 크기 조정

# Enable Compute Optimizer at account level
aws compute-optimizer update-enrollment-status \
  --status Active

# Get EC2 instance recommendations
aws compute-optimizer get-ec2-instance-recommendations \
  --filters Name=Finding,Values=OVER_PROVISIONED \
  --query 'instanceRecommendations[*].{Instance: instanceArn, Current: currentInstanceType, Recommended: recommendationOptions[0].instanceType}' \
  --output table

시나리오 4: 일괄 처리를 위한 Spot Instances

# AWS Batch compute environment with Spot instances
aws batch create-compute-environment \
  --compute-environment-name spot-genomics \
  --type MANAGED \
  --state ENABLED \
  --compute-resources '{
    "type": "SPOT",
    "bidPercentage": 60,
    "minvCpus": 0,
    "maxvCpus": 256,
    "instanceTypes": ["optimal"],
    "subnets": ["subnet-1a", "subnet-1b"],
    "securityGroupIds": ["sg-batch"],
    "instanceRole": "arn:aws:iam::123456789012:instance-profile/ecsInstanceRole",
    "spotIamFleetRole": "arn:aws:iam::123456789012:role/AmazonEC2SpotFleetRole"
  }' \
  --service-role arn:aws:iam::123456789012:role/AWSBatchServiceRole

시나리오 5: 가변 트래픽을 위한 DynamoDB On-Demand

# Switch existing DynamoDB table to On-Demand mode
aws dynamodb update-table \
  --table-name Leaderboard \
  --billing-mode PAY_PER_REQUEST

# Verify the change
aws dynamodb describe-table \
  --table-name Leaderboard \
  --query 'Table.BillingModeSummary.BillingMode'

시나리오 6: 예측할 수 없는 액세스를 위한 S3 Intelligent-Tiering

# Move objects to Intelligent-Tiering via lifecycle policy
aws s3api put-bucket-lifecycle-configuration \
  --bucket user-images-bucket \
  --lifecycle-configuration '{
    "Rules": [{
      "ID": "AutoTier",
      "Status": "Enabled",
      "Filter": {},
      "Transitions": [{
        "Days": 0,
        "StorageClass": "INTELLIGENT_TIERING"
      }]
    }]
  }'

시나리오 7: Athena와 Redshift의 절충안

# Athena cost estimate for 10 queries/week:
# Assume each query scans 5 GB of Parquet data
# 10 queries x 5 GB = 50 GB / week = 200 GB / month
# Athena cost: 200 GB x $0.005/GB = $1.00 / month
#
# Redshift dc2.large cost: $0.25/hr x 24hr x 30days = $180/month
#
# For 10 queries/week -> Athena saves $179/month
# Breakeven: when queries scan >36 TB/month or concurrency >50/day -> use Redshift

시나리오 8: EBS 볼륨 유형 선택

# Create io2 Block Express volume with 64,000 IOPS
aws ec2 create-volume \
  --volume-type io2 \
  --size 500 \
  --iops 64000 \
  --availability-zone us-east-1a \
  --encrypted

# EBS volume type IOPS limits summary:
# gp3: up to 16,000 IOPS (default 3,000, configurable)
# io1: up to 64,000 IOPS (on Nitro instances)
# io2 Block Express: up to 256,000 IOPS
# st1 (throughput HDD): no IOPS focus, max 500 MB/s throughput
# sc1 (cold HDD): lowest cost, 250 MB/s max, rarely accessed data

시나리오 9: 지속적인 워크로드를 위한 Reserved Instances

# Reserved Instance purchase decision matrix:
# On-Demand:          No commitment, highest price, any workload
# 1-yr RI (All Up):   40% discount, 1-yr commitment, specific instance type
# 3-yr RI (All Up):   60-72% discount, 3-yr commitment, best for stable workloads
# Compute Savings Plan: 66% max discount, flexible instance family/size/Region
# EC2 Spot:           90% discount, interruptible, batch/stateless only
#
# Rule: if usage > 70% of the time for >1 year -> buy RI or Savings Plan
# Rule: if usage < 50% -> stick with On-Demand
# Rule: if usage pattern is steady 3yr -> 3yr RI All Upfront maximises savings

시나리오 10: 가변 트래픽에서 Lambda와 EC2 비용 비교

# Lambda vs EC2 cost rough break-even calculation:
# Lambda costs: $0.20 per 1M requests + $0.0000166667 per GB-second
# At 128 MB memory, 100ms duration:
#   GB-seconds per request = 0.128 GB x 0.1s = 0.0128 GB-s
#   Cost per request = $0.0000166667 x 0.0128 = $0.000000213 compute
#   + $0.0000002 request fee = $0.000000413 total per request
#
# EC2 t3.micro: $0.0104/hr x 720 hrs = $7.49/month
# Break-even: $7.49 / $0.000000413 = ~18 million requests/month
# Below 18M requests/month -> Lambda cheaper
# Above 18M requests/month -> EC2 cheaper (if utilisation is high)

시나리오 11: 동적 API를 위한 Global Accelerator

# Create a Global Accelerator for an ALB
aws globalaccelerator create-accelerator \
  --name my-api-accelerator \
  --ip-address-type IPV4 \
  --enabled

# Add a listener and endpoint group pointing to ALB
aws globalaccelerator create-listener \
  --accelerator-arn arn:aws:globalaccelerator::123:accelerator/abc \
  --protocol TCP \
  --port-ranges '[{"FromPort": 443, "ToPort": 443}]'

# Endpoint group in us-east-1 with ALB
aws globalaccelerator create-endpoint-group \
  --listener-arn arn:aws:globalaccelerator::123:listener/xyz \
  --endpoint-group-region us-east-1 \
  --endpoint-configurations '[{"EndpointId": "arn:aws:elasticloadbalancing:...", "Weight": 100}]'

빠른 확인

레슨 요약