Database · 54 checks
Idle databases are the silent budget killer.
Databases run around the clock (and bill around the clock) even when nobody's querying them. Dev instances left on Multi-AZ, read replicas no one reads, provisioned throughput sized for a load test: it all adds up fast.
Graymole covers RDS, Aurora, DynamoDB, ElastiCache, Redshift, and OpenSearch on AWS; Cloud SQL, Memorystore, and BigQuery on GCP; and the Azure database services, checking utilization and configuration to separate must-keep from pure waste.
Where the money is
- Idle non-prod databases are prime candidates to stop, schedule off-hours, or downsize.
- Multi-AZ on dev/staging doubles the cost for high-availability those tiers don't need.
- DynamoDB provisioned-vs-on-demand mismatches and sparse indexes overpay every hour.
- Generation migrations (e.g. Redshift DC2→RA3, older ElastiCache nodes) cut rates substantially.
2×
A dev RDS instance left on Multi-AZ pays double for failover protection a non-production database doesn't need.
What we check
A representative sample: Database has 54 checks in total across AWS, GCP, and Azure.
RDS idle instances
Databases with near-zero connections still fully billed.
Idle Cloud SQL instances
GCP databases nobody queries, billing around the clock.
Idle Memorystore Redis
GCP Redis instances with no meaningful load.
BigQuery idle slot reservations
Reserved slots sitting unused while you pay for them.
RDS Multi-AZ on dev/staging
Paying 2× for HA non-prod doesn't require.
Aurora unused read replica
Replicas serving no read traffic.
DynamoDB provisioned waste
Provisioned capacity far above consumed.
ElastiCache idle clusters
Redis/Memcached clusters with no meaningful load.
Redshift DC2 to RA3 migration
Newer node type with better price/performance.
RDS Extended Support surcharge
Paying legacy-version surcharges you can avoid.
See your database savings
Connect a read-only role and run a free scan. Database findings come priced in real dollars with a fix for each.
Start scanning free