Question

A financial services company runs a reporting platform on an Amazon Aurora MySQL cluster with one writer and four reader replicas. Two of the readers are large, memory-optimized instances dedicated to heavy analytical queries, while the other two smaller readers serve the customer-facing dashboard. Currently the application uses the built-in reader endpoint, which round-robins connections across all four readers, causing the dashboard to occasionally hit the large analytics instances and analytical queries to land on the small instances, resulting in inconsistent performance. The team wants to ensure dashboard traffic only uses the two small readers and analytics traffic only uses the two large readers, without managing connection logic in application code. What should the solutions architect recommend?

Accepted Answer

Create two Aurora custom endpoints, one grouping the two small readers for the dashboard and one grouping the two large readers for analytics, and point each workload at its respective endpoint.

Answer

Configure Amazon RDS Proxy in front of the cluster and rely on its connection pooling to route queries to the appropriate reader instances.

Answer

Split the cluster into two separate Aurora clusters and use cross-cluster replication to keep the analytics cluster synchronized with the dashboard cluster.

Answer

Enable Aurora Auto Scaling on the reader tier so that queries are automatically distributed to correctly sized replicas based on load.