From 0 to Limitless, the 15-year odyssey of Amazon relational database product line
It’s another annual Amazon re:Invent, and the most important release of the relational database product line is Amazon Aurora Limitless Database. In the Keynote of AWS Senior Vice President Peter DeSantis, he also spent nearly half of the time going over the history of the Amazon relational database.
2009 - RDS
O to 1, lifting the vanilla MySQL and PostgreSQL to the cloud.
2014 - Aurora
Unveiling Aurora, rebuild the storage engine based on the internal log architecture (code name Grover). Aurora greatly improves the performance and availability while maintaining the full MySQL and PostgreSQL compatibility.
2018 - Aurora Serverless
Introducing Aurora Serverless, leveraging the database-optimized virtualization technology (codenamed Caspian), to offer the seamless scale up/down.
2023 - Aurora Limitless
The new Aurora Limitless, the scale-out distributed database by implementing the ultra-low latency clock synchronization.
Architecture-wise, Aurora Limitless resembles Google Spanner, both are distributed databases (NewSQL). The most difficult point of a distributed database is to implement high-performance distributed transactions. Aurora Limitless also adopts a solution similar to Spanner's TrueTime. I expect AWS to reveal details about its compatibility with native PostgreSQL and the performance benchmark shortly.
Speaking of database compatibility, this time Aurora Limitless launches PostgreSQL support first instead of MySQL. I have two hypotheses. One is that PostgreSQL codebase is easier to adapt to the Aurora Limitless architecture. Limitless requires a Router component that parses SQL, and the PostgreSQL server layer code is easier to work with; the other is that the PostgreSQL adoption has caught up with MySQL.
Summary
| Cloud | Singe-node Enhancement | Scale up/down Elasticity | Scale-out Infinity | Technology Breakthrough 🚀 | |
|---|---|---|---|---|---|
| 2009 - RDS | ✅ | General virtualization | |||
| 2014 - Aurora | ✅ | ✅ | Log-based architecture (Grover) | ||
| 2018 - Aurora Serverless | ✅ | ✅ | ✅ | Database-optimized virtualization (Caspian) | |
| 2023 - Aurora Limitless | ✅ | ✅ | ✅ | ✅ | Distributed clock synchronization |
After 15 years of iteration and 4 technological breakthroughs corresponding to 4 product generations, AWS relational database has reached its current form of Aurora Limitless. Speaking of the database core, Aurora Limitless is quite complete. What's left are the database development workflow challenges:
How to reduce the downtime of schema changes on large tables.
How to make database cloning instantaneous for development and testing.
How to make database development workflow like code flow and integrate with the overall CI/CD pipeline.
0 to Limitless completed, next from Limitless to Flawless. Go Aurora.
Top comments (0)