Service guide
How to reduce AWS RDS costs
Databases get sized once, cautiously, and never revisited — nobody wants to be the person who downsized the database. So RDS accumulates a specific set of quiet costs, including one most teams don’t know exists: extended support fees for old engine versions, billed per vCPU-hour since 2024.
The free scan checks engine versions for extended-support fees, utilisation for oversizing, and snapshot storage — per database, per region.
Run the free scanWhere RDS money goes quietly
- →Extended support fees. Running MySQL 5.7 or PostgreSQL 11-class versions past end-of-standard-support bills ~$0.10 per vCPU-hour in year one — about $146/month on a 2-vCPU instance, doubling in year three. It appears on the bill unannounced. Upgrading the engine removes it entirely.
- →Cautious oversizing. The db.r5.xlarge chosen "to be safe" at launch, running at 15% CPU three years later. Database steps price like EC2: each size down is 50%. Compute Optimizer covers RDS too.
- →Multi-AZ everywhere. Multi-AZ doubles the instance cost for failover you need in production. On dev, staging and reporting replicas it is double cost for zero benefit — rebuildable environments do not need standby replicas.
- →gp2 database storage. RDS storage on gp2 (~$0.115/GB-month) migrates to gp3 for ~20% less, live, same as EC2 volumes. Storage autoscaling that only ever grew is worth a look at the same time — RDS storage never shrinks itself.
- →Manual snapshots forever. Automated backups expire; manual snapshots don’t. Pre-migration and pre-upgrade snapshots from years past sit at $0.095/GB-month. A quarterly cleanup habit or a lifecycle policy ends it.
- →On-demand steady databases. A production database is the most predictable workload you own — and Reserved Instances for RDS discount it 30–50%+. Databases are the last place on-demand pricing makes sense and the most common place we find it.
Put numbers on it
Frequently asked questions
How do I know if I’m paying extended support fees?
Cost Explorer, filter by usage type containing "ExtendedSupport" — or check your engine versions against AWS’s end-of-standard-support dates. Our scan flags affected instances automatically with the per-instance monthly fee, since teams are usually paying it without knowing.
Is downsizing a production database safe?
It is a brief failover-length interruption on Multi-AZ, or a maintenance-window restart otherwise — so it needs scheduling, unlike most compute fixes. That is why the audit ranks database findings by evidence strength: only sustained low utilisation across weeks earns a downsize recommendation.
£499 fixed. Free scan first. 20%+ found or it’s free.
The scan is read-only — a role you create and delete, no keys shared — and shows your estimated monthly saving before anyone talks about money.