Versioning vs. Coordination in Distributed Databases: Coordination's Killer

This article explores the advantages of versioning over coordination mechanisms when building highly available, low-latency, and scalable distributed database systems. Through a concrete example, the author demonstrates how versioning avoids concurrency issues and scalability bottlenecks caused by locking. Versioning creates multiple versions of data, allowing concurrent transactions to access data without blocking each other, thus improving system performance and throughput. The article delves into version selection and management mechanisms, explaining how Aurora DSQL uses physical clocks to avoid coordination, ultimately achieving a high-performance and highly available distributed database system.