This article explains how to diagnose performance bottlenecks in Amazon Aurora and RDS for MySQL using CloudWatch Database Insights. It describes how to interpret average active sessions (AAS) against vCPU capacity and analyze wait events from the Performance Schema to identify specific resource contention.
Key points
Database load is measured as average active sessions (AAS), providing a clearer view of performance than raw CPU usage.
Comparing DB load against the Max vCPU line helps distinguish between resource-constrained states and specific query-level bottlenecks.
Wait events derived from the MySQL Performance Schema allow developers to correlate session activity with specific SQL statements.
Effective database tuning requires mapping wait events to underlying issues like I/O, locking, or inefficient indexing rather than simply scaling instance size.