# Amazon Aurora Amazon Aurora redesigns PostgreSQL by completely replacing the native storage engine with a purpose-built, highly distributed, log-structured storage service. ![[arch_cluster_aurora.png|256]] ## Log-Structured Storage In a traditional Postgres architecture, the database must write both the Write-Ahead Log (WAL) and the actual data pages to disk, requiring expensive operations like Checkpoints and Full-Page Writes. Aurora eliminates this. In Aurora, **the log is the database**. The compute node only sends WAL records over the network to the storage layer. The intelligent storage nodes then asynchronously "materialize" the database pages in the background by applying these log records. This drastically reduces network I/O and entirely eliminates traditional checkpointing. ## 6-Way Replication Aurora's storage spans three Availability Zones (AZs). Every single piece of data is sliced into 10GB segments, and each segment is replicated 6 times (2 copies per AZ). This continuous, highly fault-tolerant replication means Aurora can survive the total loss of an entire Availability Zone plus one additional storage node without losing data availability or durability. Read Replicas do not require their own storage; they simply act as additional stateless compute nodes querying the identical shared storage volume.