AWS Aurora DSQL: A Database Architecture Revolution from JVM to Rust

AWS announced Aurora DSQL at re:Invent, and this post delves into its development journey. To tackle the challenge of horizontally scaling writes in traditional databases, the AWS team initially used JVM but encountered severe performance bottlenecks, especially tail latency issues caused by garbage collection. They ultimately decided to rewrite the data plane in Rust, which significantly improved performance and reduced tail latency. Furthermore, the team also migrated the control plane to Rust, avoiding the complexities of multiple languages. This project demonstrates that questioning existing solutions and experimenting with new technologies (even costly language migrations) can yield significant rewards.
Read more