Amazon Aurora DSQL: Scalable, Distributed Relational Databases
summary
Claim: Amazon Aurora DSQL gives virtually unlimited scalability with single‑region HA ~99.99% and active multi‑region five‑nines availability, currently in public preview. Architecture builds per‑transaction microVM query processors (Firecracker) that use snapshot isolation and optimistic concurrency; commits coordinate via sharded adjudicator and journal components; durability is acknowledged at the journal, not storage. Measured tradeoff: normal reads/writes stay local and fast, cross‑region commit latency is 15–100+ ms; conflict aborts require client retry.
AI-written summary · about
chapters (13)
- 0:00Introduction and Speaker Introduction
- 0:14Current Relational Database Challenges
- 2:24Introducing Amazon Aurora DS SQL
- 2:56Architecture and Scalability of Aurora DS SQL
- 3:47Serverless Design and Ease of Use
- 4:44Single Region vs Multi-Region Clusters
- 8:05Components of Aurora DS SQL
- 9:58Transaction Management and Query Processing
- 20:50Conflict Resolution and Adjudicator Role
- 25:41Durability and Journal System
- 28:41Optimistic Concurrency Control
- 30:10Active Architecture and Failure Handling
- 34:01Conclusion and Further Resources
related talks
28mGet started with Amazon Aurora DSQL
55mAmazon Aurora: A deep-dive into what's new
43mKeep calm and dive deep into newest database innovations
20mMaximise database availability
31mDeep dive on cross-Region DR with Amazon Aurora global databases
21mSave, Grow and Innovate using AWS Databases
43mDeep dive into Amazon Aurora and its innovations
39mAWS Aurora