AWS RDS vs a managed database provider: the TCO nobody calculates
TL;DR: RDS charges a premium on compute and storage and still leaves workload-level operations (tuning, schema, capacity, incident response above the engine) with your team. Self-managed EC2 looks cheapest and usually costs the most once engineering time is priced. An MSP running databases in your own accounts combines EC2-level infrastructure pricing with an operations team, at a fixed fee. Which wins depends on your scale and how much your engineers' time is worth.
"Just put it on RDS" is the default answer to database operations in AWS shops, and for good reason: it removes real work.
It is also where most managed PostgreSQL and MySQL estates actually live, which makes it the right baseline for this comparison. But RDS is not a managed database service in the sense most teams assume.
It manages the engine, not your workload. The total cost comparison between RDS, self-managing on EC2, and a managed service provider is rarely done honestly, because two of the four cost categories are invisible on any invoice.
The four cost categories
1. Infrastructure.
RDS carries roughly a 20-30% managed-service premium over an equivalent EC2 instance, and on the relational engines PostgreSQL typically adds about 10% over MySQL for the same compute.
Storage and backups stack on top: RDS gp3 storage starts around $0.115/GB-month, with the first 100 GB of backup free and data transfer out at $0.09/GB. Multi-AZ doubles compute because the standby bills at the primary's rate. And a 2026 gotcha worth flagging: extended-support rates for MySQL 5.7 and PostgreSQL 11 doubled on March 1, 2026, so older-version estates got quietly more expensive.
2. The meter items.
Egress, NAT processing, backup storage beyond the free window, cross-region replication, snapshots. Industry analyses of cloud bill overruns consistently find these auxiliary charges, not core compute, drive the surprise. One terabyte of outbound traffic through a NAT gateway costs roughly $135 combined processing and egress.
3. Engineering time.
The category self-managed advocates undercount. At a typical US database engineer rate near $50/hour, 20 hours a month of care is about $1,000/month, more than many small RDS bills. And RDS does not reduce this to zero: query tuning, schema evolution, capacity planning, upgrade testing, and incident response above the engine remain your work. RDS manages the database process, not the database.
4. Incident cost.
The category nobody budgets. An outage's cost is your revenue exposure times duration, and it lands regardless of which model you chose. What differs is who is working the incident at random times and how fast. RDS gives you an engine restart and a support case; your team owns the rest.
Where each model actually wins
RDS wins for small numbers of standard databases, no dedicated database capacity on the team, and workloads whose growth is modest. The premium buys real convenience, and at small scale the premium is small in absolute terms.
Self-managed wins almost never on true TCO, unless you already employ database engineers with spare capacity, in which case their marginal cost is genuinely low. The published math is consistent: the infrastructure savings evaporate the first month your team spends 20+ hours on database care.
An MSP in your accounts wins when three conditions hold: the databases are business-critical (incident cost is real), the estate is large or growing (the RDS premium and meter items compound), and your engineers have better things to do than operate infrastructure. You pay EC2-level infrastructure prices in your own accounts, keep full cost visibility, and buy the operations as a fixed fee that does not grow with your data.
The worked comparison to run for your own estate
Take your last three months of AWS database spend and split it: instance and storage cost, then every meter item (egress, backups, snapshots, NAT, cross-region). Add engineering hours spent on database work at loaded cost, including the incident weeks. Now price the same infrastructure at EC2/EBS rates in your accounts plus a fixed MSP fee. For estates above roughly $10-15k/month of database spend, the comparison usually gets uncomfortable for the status quo.
FAQ
Is RDS a fully managed database service?
RDS manages the engine: provisioning, patching, backups, failover. Workload-level operations (query tuning, schema management, capacity planning, and incident response above the engine) remain the customer's responsibility. That distinction drives most RDS TCO surprises.
Is self-managing databases on EC2 cheaper than RDS?
On infrastructure sticker price, substantially. On total cost, rarely: the engineering hours required to run production databases reliably typically exceed the infrastructure savings unless you already employ underutilized database specialists.
Can a managed service provider run databases inside our AWS account?
Yes. In-your-account operation means you keep AWS pricing, billing visibility, data ownership, and compliance control, while the provider operates under your policies with committed SLAs.
When does an MSP become cheaper than RDS?
As estate size, growth rate, and criticality rise. The RDS premium, Multi-AZ doubling, and meter items scale with usage; a fixed operations fee does not. Run the four-category comparison above on your own numbers.
