Amazon Aurora - Design for High Throughput Cloud-Native Relational DBs
Explore the architecture and design considerations of Amazon Aurora, a cloud-native relational database service for OLTP workloads, in this conference talk from PASS Data Community Summit. Delve into how Aurora addresses the central constraint of network traffic in high throughput data processing by implementing a novel architecture that pushes redo processing to a multi-tenant scale-out storage service. Learn about the benefits of this approach, including fast crash recovery, failover to replicas without data loss, and fault-tolerant, self-healing storage. Discover how Aurora avoids distributed consensus algorithms in most scenarios by establishing invariants and leveraging local transient state, resulting in improved performance, reduced variability, and lower costs. Gain insights into the future of cloud-native relational databases and their role in modern data processing architectures.