Comparison

Updated October 25, 2026

· ~2 min read

DynamoDB vs Aurora PostgreSQL vs Aurora DSQL (2026): Key-Value, Real Postgres or Serverless SQL

Aurora DSQL
DynamoDB
Aurora PostgreSQL
Databases
Serverless

AWS now has three serious answers for a new application's primary database. DynamoDB is key-value with predictable latency at any scale. Aurora PostgreSQL is Postgres with Aurora's storage underneath. Aurora DSQL, generally available since May 2025, is a distributed, serverless, PostgreSQL-compatible database with active-active multi-Region writes and nothing to size.

DSQL launched without foreign keys, sequences or identity columns, which ruled it out for most ORMs and most migrations. Sequences and identity columns arrived in February 2026 and foreign keys in August. That makes it a real contender. It is still not Postgres, and the differences decide where it fits.

Side by side

Aurora DSQL

DynamoDB

Aurora PostgreSQL

Data model

Relational, PostgreSQL-compatible SQL

Key-value and document; queries by key and index

Relational, full PostgreSQL

Capacity

None to configure; scales and bills on use

On-demand or provisioned capacity

Instances, or Serverless v2 in ACUs

Multi-Region writes

Active-active, strongly consistent

Global tables, multi-active; multi-Region strong consistency available

Global Database: one writer Region, replicas elsewhere

Concurrency

Optimistic: conflicts fail at commit and must be retried

Conditional writes and transactions (up to 100 items)

Locking, MVCC, any isolation level

Transaction size

Up to 3,000 modified rows

Up to 100 items

Practically unbounded

Postgres features

No triggers, PL/pgSQL, temp tables or TRUNCATE; one database per cluster; Repeatable Read only

Not applicable

Everything, including extensions

Authentication

IAM-generated tokens

IAM

Passwords, IAM or Secrets Manager

Pricing

Unit

US East price

Free tier

Aurora DSQL

Distributed Processing Units (compute, I/O, CDC) and storage

$8 per million DPU; $0.33/GB-month

100,000 DPU and 1 GB a month

DynamoDB on-demand

Read and write request units, storage

$0.125 per million reads; $0.625 per million writes; $0.25/GB-month

25 GB storage

Aurora Serverless v2

ACU-hours (about 2 GiB of memory each), storage, I/O

$0.12 per ACU-hour; $0.10/GB-month; $0.20 per million I/Os (Standard)

None

What optimistic concurrency means for your code

DSQL does not lock rows. Two transactions that touch the same row both proceed, and the second to commit gets a serialization error. That removes deadlocks and lets the database spread across nodes and Regions, and it moves work into the application. Every write path needs retry logic, and the transaction has to be idempotent so a retry is safe. Hot rows (a global counter, an inventory row everyone decrements) become retry storms. AWS's guidance is to use random keys and spread updates across the key range. Sequences are supported, but a sequential key concentrates writes in one place, which is the pattern OCC handles worst.

Which one

Question 1 of 3

Do you need full PostgreSQL: extensions, triggers, stored procedures, long or large transactions?

Our verdict

**Aurora PostgreSQL:** you want Postgres: extensions, PL/pgSQL, triggers, long transactions, pgvector, and any ORM or tool without caveats. The default for most products. **DynamoDB:** access patterns you can name up front, very high or very spiky traffic, single-digit-millisecond reads at any size, and a team comfortable designing for keys rather than queries. **Aurora DSQL:** new apps that want SQL with no capacity to manage, scale to zero, or active-active writes across Regions, and can live with optimistic concurrency and a 3,000-row transaction limit.

Related solutions

AWS Architecture Review

A focused, senior-level review of your current AWS architecture, with a prioritized list of what to fix and why.

1 week
$4k–$8k
View

Multi-Tenant SaaS Platform Architecture

Production-ready SaaS architecture supporting secure multi-tenant isolation and rapid scaling.

4–10 weeks
$25k–$60k
View

Multi-Region Disaster Recovery

Implement a tested disaster recovery strategy that meets your RTO and RPO, before you need it.

3–6 weeks
$15k–$40k
View

Get your database choice reviewed before you build on it

We work with startups and growing companies on AWS architecture decisions. Tell us about your situation and we will recommend the right approach.

As of October 25, 2026. Provided as is, for general information, and accurate to the best of our knowledge on the date shown. Not legal, compliance, tax, financial or other professional advice. Prices, limits and vendor services change; check current vendor documentation before relying on anything here. Terms of use.