Everything you don’t need to know about Amazon Aurora DSQL: Part 4 – DSQL components

This article details the architectural components of Amazon Aurora DSQL, a distributed active-active database system.

• Query Processor (QP) executes SQL, manages transactions, operates in Firecracker microVMs without handling durability or concurrency control
• Adjudicator system maintains consistency through lease-based mechanisms, sharded across key ranges to prevent bottlenecks
• Journal stores post-images of transactions, delegates durability responsibility, uses S3 snapshots for recovery
• Crossbar merges multiple journals into ordered sequence, distributes data to storage shards using erasure codes for tail latency reduction
• Storage layer provides long-term durability via S3 snapshots, handles reads directly from QP, implements garbage collection based on trim horizon
• Aurora DSQL deliberately avoids data caching, instead pushing operations down to minimize round trips

Aurora DSQL separates traditional database responsibilities across specialized components, enabling scalable, consistent, distributed transaction processing without single points of failure.