RDS vs Aurora vs DynamoDB — When to Choose

Intermediate·2 min read·

Fast recall

Quick revision

Core answer

Choose RDS for standard relational workloads with Spring Data JPA, Aurora for relational workloads that need higher performance, global distribution, or serverless auto-scaling, and DynamoDB for high-scale key-value or document access patterns where you can design around a single-table model.

Interview Answer2 min

AWS offers three primary managed database options for Java backend services — RDS, Aurora, and DynamoDB — and the right choice depends on whether you need relational semantics, operational simplicity, or scale-first design.

RDS — standard relational, full Hibernate support

RDS runs managed PostgreSQL, MySQL, Oracle, or SQL Server — which means Spring Data JPA, Hibernate, and JDBC work without any code changes; you choose the engine version and instance size, and AWS handles backups, patching, and failover
The limitation: RDS scales vertically — when you need more capacity, you move to a larger instance class; there's a ceiling, and the migration requires a brief downtime

Aurora — relational with cloud-native storage

Aurora separates compute from storage — the storage layer replicates across six copies in three Availability Zones automatically, which means a storage node failure doesn't cause data loss and read replicas can be added with no additional storage cost
Aurora PostgreSQL is compatible with PostgreSQL, so Spring Data JPA and Hibernate work identically; the difference is Aurora's storage auto-scales from 10GB to 128TB and multi-master configurations allow write scaling
Aurora Serverless v2 scales compute capacity in fine-grained increments between a minimum and maximum — which means you pay for what you use and idle services scale close to zero

DynamoDB — key-value, single-digit millisecond at any scale

DynamoDB is a NoSQL key-value and document store that handles any amount of traffic by partitioning data horizontally — you define a partition key and optionally a sort key; queries are limited to the key structure
Spring Data DynamoDB or the Enhanced DynamoDB Client maps Java objects to DynamoDB items; there's no JPA, no SQL, and no joins — cross-entity queries require pre-computed indexes (GSI) or denormalized data
Choose DynamoDB when: workloads are simple key lookups or range scans on the primary key, traffic is highly variable, or the dataset is too large for a relational engine

Interview-ready guide

Interview answer

RDS — standard relational, full Hibernate support

RDS runs managed PostgreSQL, MySQL, Oracle, or SQL Server — which means Spring Data JPA, Hibernate, and JDBC work without any code changes; you choose the engine version and instance size, and AWS handles backups, patching, and failover
The limitation: RDS scales vertically — when you need more capacity, you move to a larger instance class; there's a ceiling, and the migration requires a brief downtime

Aurora — relational with cloud-native storage

Aurora separates compute from storage — the storage layer replicates across six copies in three Availability Zones automatically, which means a storage node failure doesn't cause data loss and read replicas can be added with no additional storage cost
Aurora PostgreSQL is compatible with PostgreSQL, so Spring Data JPA and Hibernate work identically; the difference is Aurora's storage auto-scales from 10GB to 128TB and multi-master configurations allow write scaling
Aurora Serverless v2 scales compute capacity in fine-grained increments between a minimum and maximum — which means you pay for what you use and idle services scale close to zero

DynamoDB — key-value, single-digit millisecond at any scale

DynamoDB is a NoSQL key-value and document store that handles any amount of traffic by partitioning data horizontally — you define a partition key and optionally a sort key; queries are limited to the key structure
Spring Data DynamoDB or the Enhanced DynamoDB Client maps Java objects to DynamoDB items; there's no JPA, no SQL, and no joins — cross-entity queries require pre-computed indexes (GSI) or denormalized data
Choose DynamoDB when: workloads are simple key lookups or range scans on the primary key, traffic is highly variable, or the dataset is too large for a relational engine

03 · Deep dive