A single db.r6g.large instance running PostgreSQL on AWS RDS on-demand costs $0.26 per hour in us-east-1. Over 12 months, that’s $2,277.60 before you add storage, I/O, and backup costs. Whether that’s acceptable depends entirely on your workload pattern, your tolerance for upfront commitment, and whether AWS is even the right database host for your use case.
This article models 12-month TCO across four options: RDS on-demand, RDS 1-year reserved, RDS 3-year reserved, and Aurora Serverless v2. Then it benchmarks all three against DigitalOcean Managed PostgreSQL at the same workload tier. All prices are pulled from public pricing pages and reflect us-east-1 / AWS standard pricing as of 2025.
Use the InfraTally DB Cost Calculator to run your own numbers before committing to anything.
To make this comparison meaningful, every option is modeled against the same workload:
This is a realistic spec for a production application serving a few hundred concurrent users with a normalized OLTP schema.
The db.r6g.large instance (2 vCPU, 16 GB RAM) is the workhorse instance for this workload class.
On-demand rate: $0.26/hr (us-east-1, PostgreSQL, Single-AZ)
| Cost Component | Monthly | 12-Month Total |
|---|---|---|
| Instance (on-demand) | $189.80 | $2,277.60 |
| Storage (100 GB gp3) | $11.50 | $138.00 |
| I/O (gp3, included in base) | $0.00 | $0.00 |
| Backup storage (100 GB) | $9.50 | $114.00 |
| Total | $210.80 | $2,529.60 |
Storage pricing: $0.115/GB-month for gp3. Backup storage above the allocated size costs $0.095/GB-month. These are published RDS rates.
On-demand works for unpredictable or short-lived workloads. For anything running continuously for 12+ months, you’re leaving significant money on the table.
Reserved Instances (RIs) are a capacity commitment, not a technical change. The database behaves identically. You’re just prepaying in exchange for a discount.
AWS offers three payment options: No Upfront, Partial Upfront, and All Upfront. For a db.r6g.large in us-east-1:
1-Year Reserved (All Upfront): $0.161/hr effective rate 3-Year Reserved (All Upfront): $0.113/hr effective rate
| Option | Effective Hourly Rate | 12-Month Instance Cost | 12-Month Total TCO |
|---|---|---|---|
| On-Demand | $0.260/hr | $2,277.60 | $2,529.60 |
| 1-Yr RI (All Upfront) | $0.161/hr | $1,410.36 | $1,662.36 |
| 1-Yr RI (No Upfront) | $0.177/hr | $1,549.20 | $1,801.20 |
| 3-Yr RI (All Upfront) | $0.113/hr | $989.88 | $1,241.88 |
Storage and backup costs ($252/yr) are added to each row. I/O is zero because gp3 includes baseline throughput.
The 1-year All Upfront RI saves $867.24 over 12 months compared to on-demand. That’s a 34% reduction. The breakeven point on paying upfront vs no-upfront is around month 3.
The 3-year RI is compelling on paper, but it requires confidence that the workload and instance class won’t change for 36 months. Instance families shift, workloads scale, and being locked into a db.r6g.large while Aurora becomes a better fit is a real operational cost even if it doesn’t show on an invoice.
Convertible RIs partially solve this by allowing you to exchange for an RI of equal or greater value. For most teams doing workload migrations, Convertible 1-year is the pragmatic middle ground.
Aurora Serverless v2 bills in Aurora Capacity Units (ACUs). Each ACU is approximately 2 GB of memory. Pricing in us-east-1:
For a workload that needs 8 ACUs at peak (roughly equivalent to the 16 GB RAM baseline) but drops to 2 ACUs overnight and 0.5 ACUs on weekends, the math changes significantly.
Aurora Serverless v2 workload model:
| Time Window | Hours/Month | ACUs | ACU-Hours | Monthly Cost |
|---|---|---|---|---|
| Business hours (8hr/day, 22 days) | 176 | 8 | 1,408 | $168.96 |
| Off-hours weekdays | 176 | 2 | 352 | $42.24 |
| Weekend (48hr x 4) | 192 | 0.5 | 96 | $11.52 |
| Total | 544 | 1,856 | $222.72 |
Add Aurora storage at $0.10/GB-month (100 GB = $10.00) and I/O at $0.20 per 1 million requests (approximately $6.00/month at 1M req/day). Total: $238.72/month, or $2,864.64/year.
At this workload pattern, Aurora Serverless v2 is more expensive than on-demand RDS and significantly more expensive than a 1-year RI. Aurora Serverless v2 only wins when utilization drops sharply and frequently. A workload running at 60%+ capacity most of the time will always pay more per ACU than a provisioned instance.
Aurora Serverless v2 is the right pick when you have genuinely spiky traffic (think nightly batch jobs that run 2 hours and are silent the other 22) or when you’re prototyping and want to avoid the RI commitment. It is not a cost-reduction tool for steady-state workloads.
One hard Aurora overhead to factor in: Aurora always requires a minimum of 2 instances for Multi-AZ in a cluster. The single-writer / zero readers Serverless v2 model is supported, but check your cluster configuration before assuming you’re only paying for one instance’s ACUs.
DigitalOcean Managed PostgreSQL takes a flat-rate approach. No reserved commitments, no ACU math, no I/O pricing. The closest plan to our baseline (2 vCPU, 16 GB RAM) is the General Purpose 4 vCPU / 16 GB node:
$200/month (as published on DigitalOcean’s managed database pricing page, 2025)
Storage is included up to the plan limit (typically 200 GB at this tier), which covers the 100 GB baseline with headroom.
| Option | Monthly Cost | 12-Month TCO |
|---|---|---|
| RDS On-Demand (db.r6g.large) | $210.80 | $2,529.60 |
| RDS 1-Yr RI All Upfront | $138.53 | $1,662.36 |
| RDS 3-Yr RI All Upfront | $103.49 | $1,241.88 |
| Aurora Serverless v2 (variable) | $238.72 | $2,864.64 |
| DigitalOcean Managed PG | $200.00 | $2,400.00 |
DigitalOcean sits between RDS on-demand and the 1-year RI in annual cost. No upfront payment, no commitment, PostgreSQL-native feature support, and a predictable invoice.
Where DigitalOcean wins: teams that want a flat billing model, don’t want to think about I/O pricing, and are not already deep in the AWS ecosystem. The managed PostgreSQL product includes automated failover, daily backups, connection pooling via PgBouncer, and end-to-end encryption. That’s a complete production setup at a fixed monthly line item.
Where it loses: if your application is already on EC2, Lambda, or ECS in AWS, the inter-region network costs of running your database on DigitalOcean will eat the savings. This comparison assumes your application can co-locate with the database on DigitalOcean or that you’re doing a full infrastructure evaluation, not a hybrid migration.
For deeper AWS pricing context, see the InfraTally RDS Pricing Explained guide.
At what month does the 1-year RI (All Upfront) break even against on-demand?
Breakeven vs on-demand: Month 7 (cumulative on-demand surpasses the RI total cost)
Against DigitalOcean’s $200/month (no upfront):
If cash flow matters and upfront spend is constrained, DigitalOcean is the rational choice even knowing AWS’s RI discount is better on paper.
For a steady-state production PostgreSQL workload running 24/7, the 1-year All Upfront RDS RI on a db.r6g.large is the cheapest 12-month option by a wide margin if you’re already on AWS. The math is clear and the risk is manageable.
Aurora Serverless v2 is a pricing trap for predictable workloads. It makes sense for development environments, infrequent batch jobs, and genuinely bursty applications. Do not reach for it as a “modern” alternative to provisioned instances when your load is steady.
DigitalOcean Managed PostgreSQL is the right call when you want PostgreSQL without the AWS billing complexity, when your team isn’t committed to the AWS ecosystem, or when flat-rate predictability matters more than squeezing the last dollar out of reserved capacity pricing. At $200/month with no commitment, it’s a serious option, not a budget fallback.
Run your specific instance class, region, and storage size through the InfraTally DB Calculator to find your actual breakeven point before signing a 1-year or 3-year RI commitment.