Tair is the appropriate architecture for a frequently changing game leaderboard because Redis-compatible sorted data structures are optimized for high-throughput score updates and ranking operations. Instead of repeatedly executing expensive aggregation and sorting queries against the relational database, the application can maintain player scores in Tair and retrieve the highest-ranked members directly.

Alibaba Cloud specifically documents leaderboard implementations based on Tair ' s sorted-set capabilities. Tair ' s enhanced exZset functionality supports scalable ranking use cases and can distribute large leaderboards across multiple internal structures while preserving rank calculation.

This also protects RDS from high-frequency scoreboard queries. RDS remains the durable transactional database, while Tair operates as the low-latency serving and ranking layer for frequently accessed score data.

CDN caching is unsuitable because the leaderboard changes continually and requires near-real-time freshness. A second relational PostgreSQL database would duplicate database infrastructure without providing the specialized in-memory ranking semantics required. An RDS read replica can reduce pressure on the primary database, but calculating and sorting rankings through SQL for every refresh remains less efficient than maintaining rankings directly in an in-memory sorted structure.

Study Guide reference: Building Highly Available, Performant Cloud Architecture — Tair, caching, low-latency architectures, and database performance optimization.

==================================================