AWS RDS and Aurora. The 2026 database bill levers.

AWS RDS and Aurora spend bends on four levers: Reserved Instance coverage, Aurora configuration choice, commercial engine licensing, and how the database line feeds your EDP commit. None of them require a migration.

What actually moves the RDS and Aurora bill?

Four levers move the bill without a migration: Reserved Instance coverage, Aurora configuration, engine licensing mode, and instance family. RDS pricing is published per instance hour, so every lever is checkable against list.

Savings Plans do not apply to RDS. That surprises teams who assume their compute Savings Plan blankets the database line. It does not, and the gap is usually the single largest miss we find.

Why is RI coverage the first lever to pull?

Because it is pure price, zero engineering. RDS Reserved Instances discount steady state instances 25 to 45 percent against on demand, and a coverage review takes a week. Start there before any architectural debate.

How do you keep Aurora from outgrowing its budget?

Aurora bills compute, storage, and I/O separately, and the I/O line is the one that surprises. Aurora offers a standard configuration and an I/O optimized configuration that folds I/O into the instance price.

The crossover is predictable: when I/O charges pass roughly a quarter of your Aurora spend, the I/O optimized configuration wins. In our file the switch cut Aurora bills 25 to 40 percent on transaction heavy workloads.

• Serverless v2: ACU based billing fits spiky workloads, but a floor set too high erases the benefit.
• Storage growth: Aurora storage scales automatically, so archive and partition discipline is a cost control, not housekeeping.
• Read replicas: each replica is billed compute; review replica counts against actual read traffic quarterly.

When does Serverless v2 beat provisioned capacity?

When the workload is genuinely spiky and idles more than half the day. A steady workload on Serverless v2 with a high ACU floor costs more than a reserved provisioned instance doing the same work.

What does engine choice do to the licensing math?

Commercial engines carry the licensing question. License included RDS for Oracle or SQL Server reprices the license into the hourly rate, while BYOL consumes licenses you already own. Run both models against your shelf before renewal.

• License included: clean exit from vendor audits on those workloads, but you pay the license again every hour.
• BYOL: consumes owned licenses and support contracts, but keeps you inside the vendor's audit perimeter.
• Open source engines: MySQL and PostgreSQL on Graviton instances are the lowest cost stable state on the platform.

Should you migrate engines to cut the bill?

Only after the four commercial levers are exhausted. An engine migration is a project with risk; an RI purchase is a purchase order. Sequence accordingly.

How does RDS spend play into the EDP negotiation?

RDS and Aurora spend draws down an Enterprise Discount Program commit like any other AWS service, so the database growth curve belongs in your commit model. Overcommitting on the assumption of unmanaged database growth donates the optimization upside to AWS.

1. Build the 3 year database forecast after rightsizing, not before.
2. Hold RI purchases and the EDP discussion together; both reprice the same line.
3. Use the optimized run rate as the commit base and keep the growth case as upside.
4. Ask for database heavy workload incentives in the EDP if you are migrating commercial engines.
5. Rebaseline at every EDP renewal; the database line moves faster than the estate average.

What do database heavy estates get in EDP terms?

Estates where databases pass roughly a third of total spend have credibly pulled migration credits and service specific incentives into EDP renewals. The lever is a documented alternative, even a partial one.

Where the common advice on RDS and Aurora cost is wrong

The standard advice is to start with an engine migration to open source and treat everything else as noise. We disagree. In roughly 20 to 30 AWS database estates Morten Andersen benchmarked in 2024 to 2025, the unglamorous levers, Reserved Instance coverage, the Aurora I/O optimized switch, and nonprod cleanup, recovered 25 to 45 percent of the line in under a quarter with no project risk. Engine migrations took 9 to 18 months and stalled half the time. The buyer side move is to pull the pricing levers first and let the migration case stand on strategy, not savings it may never deliver.

What the engagement data shows

Three cuts of our advisory engagement file frame the size of the opportunity.

What to do next

Five moves turn this analysis into a lower invoice on the next renewal.

A sequence you can run this quarter

1. Pull 90 days of RDS and Aurora usage and tag steady state versus spiky workloads.
2. Buy or true up Reserved Instances to 80 plus percent coverage on steady state production.
3. Compare Aurora standard versus I/O optimized configuration on actual I/O spend.
4. Rightsize and schedule nonproduction instances before forecasting growth.
5. Run BYOL versus license included math on every commercial engine instance.
6. Take the optimized run rate into the next EDP sizing discussion.

Related advisory: Buyer side AWS contract negotiation support across EDP, PPA, and enterprise support plans.

White Paper · AWS

AWS RDS and Aurora Negotiation

Seven buyer side levers that cut AWS RDS and Aurora cost: instance right sizing, Reserved Instances, Aurora Serverless v2, and the EDP overlap. Read it free.

Read the white paper

Frequently asked questions

Do AWS Savings Plans cover RDS and Aurora?

No. Compute Savings Plans cover EC2, Fargate, and Lambda, not RDS. Database instances discount through RDS Reserved Instances, which is why RI coverage is the first lever on the database line.

How much do RDS Reserved Instances save?

Typically 25 to 45 percent against on demand for 1 and 3 year terms depending on engine and payment option. Steady state production databases justify 80 plus percent coverage in most estates we benchmark.

When is Aurora I/O optimized cheaper than standard?

When I/O charges pass roughly 25 percent of your Aurora spend. Transaction heavy workloads in our 2024 to 2025 file cut 25 to 40 percent off the Aurora line by switching configuration.

Is BYOL or license included better for Oracle and SQL Server on RDS?

It depends on your shelf. License included reprices the license hourly but simplifies audits; BYOL consumes licenses and support you already pay for. Run both models against owned entitlements before renewal.

Does RDS spend count toward an AWS EDP commitment?

Yes. RDS and Aurora draw down the commit like any AWS service. Size the EDP on the optimized database run rate, not the unmanaged growth curve, or the discount funds waste.

Is migrating to open source engines worth it for cost alone?

Usually not as the first move. Pricing levers recover 25 to 45 percent in a quarter with no project risk; engine migrations took 9 to 18 months in our file and stalled half the time. Migrate for strategy, price first.