Real scan — anonymized data

3 DynamoDB tables on Pay-Per-Request.
$935/mo more than Provisioned mode.

We scanned a production SaaS platform across 4 AWS regions. The DynamoDB billing mode alone accounted for 80% of the waste. Add Lambda and CloudWatch, and the total hit $1,168/mo.

By Mohamed Shehabeldin, Founder of CostPatrol

$1,168/mo in waste. Found in under 2 minutes.

3 DynamoDB tables were running on Pay-Per-Request mode. Their traffic was steady enough for Provisioned capacity with Auto Scaling. That billing mode mismatch cost $935/mo. CostPatrol flagged 26 findings automatically.

Check your account

Total savings

$1,168/mo

$1,168 - $1,293

Findings

Rules matched

Regions scanned

Every finding, with exact fix commands

26 optimization opportunities across 6 rules. Real dollar amounts. Anonymized resource names.

DynamoDB On-Demand to Provisioned Migration

DDB-002 · 3 tables in eu-west-2

$934.67/mo

svc-product-catalog-production

$730.82/mo

eu-west-2 · PAY_PER_REQUEST · Table + GSI1 · Steady traffic pattern

→ Switch to Provisioned capacity with Auto Scaling (steady traffic)

# Step 1: Switch to provisioned capacity $ aws dynamodb update-table \ --table-name svc-product-catalog-production \ --billing-mode PROVISIONED \ --provisioned-throughput ReadCapacityUnits=461,WriteCapacityUnits=137 \ --global-secondary-index-updates '[{"Update":{"IndexName":"GSI1","ProvisionedThroughput":{"ReadCapacityUnits":205,"WriteCapacityUnits":59}}}]' # Step 2: Configure table read auto scaling $ aws application-autoscaling register-scalable-target \ --service-namespace dynamodb \ --resource-id "table/svc-product-catalog-production" \ --scalable-dimension "dynamodb:table:ReadCapacityUnits" \ --min-capacity 230 --max-capacity 922 $ aws application-autoscaling put-scaling-policy \ --service-namespace dynamodb \ --resource-id "table/svc-product-catalog-production" \ --scalable-dimension "dynamodb:table:ReadCapacityUnits" \ --policy-name "read-scaling" --policy-type TargetTrackingScaling \ --target-tracking-scaling-policy-configuration \ '{"TargetValue":70,"PredefinedMetricSpecification":{"PredefinedMetricType":"DynamoDBReadCapacityUtilization"},"ScaleInCooldown":300,"ScaleOutCooldown":60}' # Step 3: Configure table write auto scaling $ aws application-autoscaling register-scalable-target \ --service-namespace dynamodb \ --resource-id "table/svc-product-catalog-production" \ --scalable-dimension "dynamodb:table:WriteCapacityUnits" \ --min-capacity 68 --max-capacity 274 # Step 4: Repeat for GSI1 (read + write auto scaling) # Same pattern: register scalable target + put scaling policy for each dimension

svc-pricing-ingress-production

$121.92/mo

eu-west-2 · PAY_PER_REQUEST · Write-heavy ingestion table

→ Switch to Provisioned capacity with Auto Scaling (steady traffic)

$ aws dynamodb update-table \ --table-name svc-pricing-ingress-production \ --billing-mode PROVISIONED \ --provisioned-throughput ReadCapacityUnits=1,WriteCapacityUnits=55 # Then configure auto scaling (same pattern as above) # Write min: 27, max: 110 | Read min: 1, max: 2

svc-event-outbox-production

$81.93/mo

eu-west-2 · PAY_PER_REQUEST · Table + GSI1 · Event-driven outbox pattern

→ Switch to Provisioned capacity with Auto Scaling (steady traffic)

$ aws dynamodb update-table \ --table-name svc-event-outbox-production \ --billing-mode PROVISIONED \ --provisioned-throughput ReadCapacityUnits=1,WriteCapacityUnits=19 \ --global-secondary-index-updates '[{"Update":{"IndexName":"GSI1","ProvisionedThroughput":{"ReadCapacityUnits":1,"WriteCapacityUnits":19}}}]' # Then configure auto scaling for table + GSI1 # Write min: 9, max: 38 | Read min: 1, max: 2 (both table and GSI)

Lambda ARM64 Migration

LAM-O001 · 14 resources in eu-west-2

$191 - $320/mo

svc-aggregator-production

$173 - $288/mo

eu-west-2 · x86_64 · nodejs20.x · 1024MB · $1,153.42/mo

→ Migrate to ARM64 (Graviton2) for 20% cost savings

$ aws lambda update-function-configuration \ --function-name svc-aggregator-production \ --architectures arm64

svc-batch-processor-production

$13 - $22/mo

eu-west-2 · x86_64 · nodejs20.x · 1024MB · $88.58/mo

→ Migrate to ARM64 (Graviton2) for 20% cost savings

$ aws lambda update-function-configuration \ --function-name svc-batch-processor-production \ --architectures arm64

and 12 more x86_64 Lambda functions — $5.56 - $9.28/mo combined

CloudWatch Metric Streams Unfiltered

CW-O003 · 1 resource in eu-west-2

$29.34/mo

Monitoring-Metric-Stream--eu-west-2

$29.34/mo

eu-west-2 · $48.90/mo · Streaming all namespaces to monitoring provider

→ Add include filters to restrict to namespaces you actually consume

# Add namespace filters to stop streaming metrics you don't use $ aws cloudwatch put-metric-stream \ --name Monitoring-Metric-Stream--eu-west-2 \ --firehose-arn arn:aws:firehose:eu-west-2:xxxxxxxxxxxx:deliverystream/Monitoring-Delivery-Stream--eu-west-2 \ --output-format json \ --include-filters '[{"Namespace":"AWS/Lambda"},{"Namespace":"AWS/DynamoDB"},{"Namespace":"AWS/ApiGateway"}]'

Lambda Memory Right-Sizing

LAM-O002 · 6 resources in eu-west-2

$14.69/mo

svc-perf-upsert-production

$6.51/mo

eu-west-2 · 1024MB allocated · 359MB peak usage · $13.02/mo

→ Reduce memory from 1024MB to 512MB (30% buffer above peak usage)

$ aws lambda update-function-configuration \ --function-name svc-perf-upsert-production \ --memory-size 512

svc-perf-sync-production

$5.63/mo

eu-west-2 · 1024MB allocated · 327MB peak usage · $11.26/mo

→ Reduce memory from 1024MB to 512MB (30% buffer above peak usage)

$ aws lambda update-function-configuration \ --function-name svc-perf-sync-production \ --memory-size 512

and 4 more over-provisioned Lambda functions — $2.55/mo combined

Unused API Gateway APIs

APIGW-O002 · 1 resource in eu-west-2

$0.00/mo

svc-api-production

$0.00/mo

eu-west-2 · REST API · Zero requests in past 30 days

→ Delete unused API Gateway (no cost savings but reduces attack surface)

$ aws apigateway delete-rest-api --rest-api-id xxxxxxxxxxxx

What AWS Cost Explorer misses

AWS Cost Explorer

✕ Shows total DynamoDB spend — can't tell you which billing mode is optimal

✕ Tells you "DynamoDB costs went up" — not that on-demand is 6.5x more expensive than provisioned

✕ Cannot analyze read/write patterns to recommend provisioned capacity

✕ No auto scaling configuration — you're on your own

CostPatrol

✓ Analyzes actual read/write patterns per table over 30 days

✓ "3 tables on Pay-Per-Request — provisioned mode saves $935/mo"

✓ Calculates exact provisioned capacity numbers from real usage data

✓ Complete auto scaling commands with min/max and target utilization

AWS tells you DynamoDB costs went up. CostPatrol tells you that 3 tables on Pay-Per-Request with steady traffic should be on Provisioned mode, saving $935/mo — and gives you the exact capacity numbers and auto scaling commands to make the switch.