Amazon Aurora is a managed relational database that's wire-compatible with MySQL and PostgreSQL but runs on a storage engine Amazon rebuilt from scratch for the cloud. Your application connects with the same drivers and SQL it would use for MySQL or Postgres; underneath, Aurora replaces the traditional single-disk storage model with a distributed, fault-tolerant layer that's the source of most of its advantages.
The distributed storage layer
A normal database writes its data and its write-ahead log to local disk. Aurora instead pushes only the log records to a purpose-built storage service that spreads data across six copies in three availability zones and reconstructs pages from the log. Because it ships log records rather than full data pages, write traffic over the network drops sharply, and because storage is replicated six ways, the database survives the loss of an entire availability zone (and can tolerate losing two copies for writes, three for reads) without data loss. Storage grows automatically up to 128 TB - no pre-provisioning.
Compute and read scaling
Aurora separates that shared storage from the database instances that read and write it. One writer instance handles writes; up to 15 read replicas share the same underlying storage, so a replica doesn't need its own copy of the data and lag is typically milliseconds rather than the seconds of traditional replication. If the writer fails, a replica is promoted in about 30 seconds. Aurora Serverless takes this further, scaling compute capacity up and down automatically with load, which suits spiky or unpredictable workloads that would otherwise force you to size for the peak.
Compatibility and migration
Because Aurora speaks the MySQL and PostgreSQL protocols, most applications move to it with no code changes - you point the connection string at the Aurora endpoint. That compatibility is the main reason teams choose it over a from-scratch cloud-native database: the operational and performance benefits of a re-architected engine, with the familiarity and ecosystem of the engine you already use. It sits alongside the standard managed MySQL and Postgres options as the higher-performance, higher-durability tier within the same RDS family.
Frequently asked questions
Is Aurora the same as MySQL or PostgreSQL? It's compatible with their protocols and SQL, so applications connect unchanged, but the storage engine underneath is Amazon's own re-architected, distributed design.
How does Aurora achieve durability? It replicates data six ways across three availability zones and reconstructs pages from log records, surviving the loss of an entire AZ without data loss.
How does read scaling work? Up to 15 read replicas share the same distributed storage rather than each holding a full copy, so replicas add read capacity with millisecond lag and fast failover.
When should I use Aurora Serverless? For spiky or unpredictable workloads where automatically scaling compute beats paying for a fixed instance sized to the peak.