RI Coverage Lift
Most environments are 40-60% RI-covered on database. 80-90% is achievable with the right portfolio strategy. See our RI strategy service.
Database spend is structurally sticky. RDS, Aurora, and DynamoDB each price differently, license differently, and reserve differently. This guide breaks down the economics and the negotiation levers that actually work.
AWS databases come in three economic shapes. RDS is provisioned and instance-based — you pay for the instance whether you use it or not. Aurora is also provisioned (or serverless) but adds I/O pricing on top, which can dwarf compute. DynamoDB is pure usage-based — pay for capacity units consumed. The pricing logic determines the negotiation lever.
| Service | Pricing Shape | Reserve Strategy |
|---|---|---|
| RDS (PostgreSQL, MySQL, MariaDB) | Instance + storage + I/O + backup | Reserved Instances (1 or 3 year) |
| RDS Oracle / SQL Server | Instance + license (LI or BYOL) | RIs + license negotiation |
| Aurora Standard | Instance + storage + per-I/O fees | Aurora RIs (instance only) |
| Aurora I/O-Optimized | Instance + storage (no per-I/O) | Aurora RIs |
| Aurora Serverless v2 | ACU-hours | Limited reserve options |
| DynamoDB On-Demand | Per-request | None (provisioned only) |
| DynamoDB Provisioned | RCU + WCU | Reserved Capacity (1 or 3 year) |
Aurora's standard pricing model charges $0.20 per million I/O operations on top of instance and storage. For OLTP workloads with high I/O, this line can exceed instance cost by 2-3x. Aurora I/O-Optimized — released in 2023 — removes the per-I/O fee in exchange for a higher per-GB storage and instance rate. The crossover point is roughly 25-30% of I/O cost relative to instance cost. Above that threshold, I/O-Optimized is meaningfully cheaper. We see customers paying for Standard when I/O-Optimized would save 30-45% on the same workload.
RDS Oracle and SQL Server come in License Included (LI) and BYOL flavors. LI bundles the license cost into the per-hour price; BYOL lets you bring an existing license. For Oracle Enterprise Edition with active support, BYOL is almost always cheaper — and frequently the only path to keep your existing Oracle support contract valid. SQL Server is more nuanced; for new workloads, LI is usually administratively simpler. This is a contract negotiation question, not a pricing one — and it should be addressed inside the EDP, not at the database team level.
DynamoDB On-Demand is convenient but expensive — roughly 6-7x the per-request cost of Provisioned at full utilization. For predictable workloads, Provisioned with Reserved Capacity delivers 50-76% off the On-Demand-equivalent price. The right answer for most workloads is Provisioned with Auto Scaling and Reserved Capacity on the floor. We have repeatedly reduced DynamoDB lines by 60%+ by simply switching modes.
Most environments are 40-60% RI-covered on database. 80-90% is achievable with the right portfolio strategy. See our RI strategy service.
Per-cluster I/O ratio analysis. We model both pricing modes against last 90 days of consumption and flip the ones that benefit.
Provisioned + Reserved Capacity on the baseline, On-Demand for spike. Most DynamoDB tables should be Provisioned.
If you have active Oracle support, BYOL on RDS is almost always cheaper. We coordinate with your Oracle reseller to validate license mobility.
Graviton-backed RDS instances (m6g, r6g, m7g) carry a 10-20% discount versus Intel and are now broadly available across engines.
Database spend rolls into EDP qualifying usage. We negotiate database-specific concessions inside EDP rather than accepting the blended tier.
500+ engagements. $340M+ client savings. We audit RDS, Aurora, and DynamoDB economics, then negotiate inside the EDP.