Real scan — anonymized data

$520/mo from one Aurora setting
most teams don't know about

We scanned a production AWS account across 12 regions. The biggest single finding was an Aurora cluster bleeding $520/mo on per-I/O charges. The fix is one AWS CLI command. Plus 92 more findings totaling $1,299-$1,350/mo across the same account.

By Mohamed Shehabeldin, Founder of CostPatrol

$1,299 - $1,350/mo across 93 findings. Found in 90 seconds.

One Aurora cluster on the wrong storage type was bleeding $520/mo on its own. The rest was idle RDS instances, stopped EC2s still paying for EBS, previous-gen instances eligible for Graviton, oversized log retention, and the usual long tail. 76 of the 93 findings sat in us-east-2 alone, not the account's primary region.

Check your account

Total savings

$1,350/mo

$1,299 - $1,350

Findings

Rules matched

Regions scanned

Every finding, with exact fix commands

93 optimization opportunities across 21 rules and 12 regions. Real dollar amounts. Anonymized resource names.

Aurora I/O-Optimized Eligibility

RDS-O009 · 1 cluster · us-east-2

$520.77/mo

prd-saas-cluster (Aurora PostgreSQL 15.12)

$520.77/mo

us-east-2 · 1 instance · Standard storage · I/O at 79.45% of cluster cost

→ Switch storage type from standard (aurora) to I/O-Optimized (aurora-iopt1). Eliminates per-I/O charges in exchange for ~30% higher instance hour rate.

Compute cost (current)$146.00/mo

I/O charges (current)$564.57/mo

Current cluster cost$710.57/mo

Compute cost (after switch)$189.80/mo

I/O charges (after switch)$0.00/mo

Monthly savings$520.77/mo

# Switch storage type to I/O-Optimized — applies immediately, no downtime $ aws rds modify-db-cluster \ --db-cluster-identifier prd-saas-cluster \ --storage-type aurora-iopt1 \ --apply-immediately # Reversible. Run with --storage-type aurora to revert.

Why most teams miss this: Aurora I/O-Optimized launched in 2023 as a non-default storage option. AWS Cost Explorer doesn't flag clusters that would benefit. AWS Trusted Advisor doesn't detect this rule. The breakeven is when I/O charges exceed 25% of cluster cost. Production clusters with even moderate write workloads frequently cross that threshold without anyone noticing because nobody is computing the comparison.

Idle RDS Instances + Unused Read Replicas

RDS-O001 + RDS-O011 · 4 resources · us-east-2

$400/mo

prd-saas-cluster-instance-1

$146/mo

us-east-2 · Aurora PostgreSQL · Avg CPU < 5%, DatabaseConnections < 5, no I/O for 7+ days

→ Stop or delete idle Aurora instance. Cluster has more capacity than the workload needs.

# Snapshot first, then stop or delete $ aws rds stop-db-instance \ --db-instance-identifier prd-saas-cluster-instance-1

prd-saas-cluster-instance-2

$146/mo

us-east-2 · Aurora PostgreSQL · Same idle pattern

→ Stop or delete

$ aws rds stop-db-instance \ --db-instance-identifier prd-saas-cluster-instance-2

prd-saas-cluster-instance-1-us-east-2c (read replica)

$146/mo

us-east-2 · Read replica · No replica traffic detected for lookback window

→ Delete unused read replica. No application is reading from it.

$ aws rds delete-db-instance \ --db-instance-identifier prd-saas-cluster-instance-1-us-east-2c \ --skip-final-snapshot

and 1 more idle RDS instance — overlap with Aurora cluster optimization deducted

Previous-Generation EC2 Instance Types

EC2-O001 · 6 resources · us-east-2

$168/mo

backend-command-executor (m4.large)

$28/mo

us-east-2 · Previous-gen m4 · Current-gen m5 is cheaper and faster

→ Migrate to current-gen instance class (m5.large or m6i.large). Better price/performance, Nitro-based.

$ aws ec2 stop-instances --instance-ids i-xxxx $ aws ec2 modify-instance-attribute --instance-id i-xxxx \ --instance-type '{"Value":"m5.large"}' $ aws ec2 start-instances --instance-ids i-xxxx

and 5 more previous-gen instances across us-east-2 — $140/mo additional savings

Stopped EC2 with Attached Storage + Graviton Migrations

EC2-O007 + EC2-O003 · 12 resources · us-east-2

$182/mo

8x stopped EC2 instances still attached to EBS

$184/mo

us-east-2 · Stopped 30+ days · EBS volumes still billing at gp2/gp3 rates

→ Snapshot required volumes, terminate instances, delete EBS volumes that aren't needed for restore.

# For each stopped instance: snapshot, terminate, then clean up EBS $ aws ec2 create-snapshot --volume-id vol-xxxx --description "pre-terminate" $ aws ec2 terminate-instances --instance-ids i-xxxx $ aws ec2 delete-volume --volume-id vol-xxxx

2x EC2 boxes tagged GitLab and Jenkins (Graviton candidates)

$48/mo

us-east-2 · m5.large + m5.xlarge · ARM64 equivalent available, ~30% cheaper

→ Migrate to m6g.large/m6g.xlarge (Graviton2). Same instance class, 30% cheaper. Verify build runner ARM compatibility first.

# Test ARM compatibility, then migrate $ aws ec2 modify-instance-attribute --instance-id i-xxxx \ --instance-type '{"Value":"m6g.large"}'

and 2 more EC2 findings — $46/mo additional savings

Long Tail: Lambda, CloudWatch Logs, NAT, EIP, Route 53

LAM-O001/2/3/4 + CWL-O001/3 + IPV4-O001 + R53-O001 + others · 70 resources

~$80/mo

24x unused Lambda functions (LAM-O003)

delete candidates

0 invocations in 30 days · Last modified 6+ months ago · Storage and config retained

→ Delete unused Lambda functions. No invocations means no value, just config drift and audit noise.

$ aws lambda delete-function --function-name FUNCTION_NAME

12x CloudWatch log groups with bad retention (CWL-O003 + CWL-O001)

$30/mo

Retention "Never expire" or 90+ days · 7-30 days would suffice for these workloads

→ Set sensible retention. Old log data costs money to store and is rarely queried.

$ aws logs put-retention-policy \ --log-group-name "/aws/lambda/FUNCTION" \ --retention-in-days 30

3x unattached Elastic IPs + 1 Route 53 health check on dead endpoint

$15/mo

EIPs: $3.65/mo each when unattached · R53 health check polling a domain that no longer responds

→ Release unattached EIPs and delete the dead health check.

$ aws ec2 release-address --allocation-id eipalloc-xxxx $ aws route53 delete-health-check --health-check-id HC_ID

and 30+ more findings across NAT Gateways, VPC Endpoints, KMS, ECR — covered in the full report

What AWS Cost Explorer misses

AWS Cost Explorer + Trusted Advisor

✕ Shows RDS spend climbing — without knowing the cluster is on the wrong storage type

✕ No detection for Aurora I/O-Optimized eligibility

✕ No surfacing that 76 of 93 findings sit outside the primary region

✕ No fix commands. You're on your own to figure out the syntax.

CostPatrol

✓ Aurora I/O-Optimized eligibility detected automatically with breakeven math

✓ Per-resource current cost vs. recommended cost vs. monthly savings

✓ Scans every region. Finds waste in regions you forgot you were using.

✓ Exact AWS CLI command to apply each fix. Copy-paste ready.

AWS shows you a bigger bill. CostPatrol shows you that one Aurora cluster is bleeding $520/mo on the wrong storage type, and 92 other resources are quietly adding another $830 on top.