A retail company hosts its relational database on a single Cloud SQL instance. During peak shopping events, write operations succeed normally, but customer-facing product searches and analytical reports experience significant query latency because they compete for compute resources on the primary instance. Which replication configuration should the database administrator implement to reduce query latency for these read-heavy workloads?

Configure Cloud SQL read replicas to offload read-only and reporting queries from the primary instance.: ### **What Cloud SQL Read Replicas Are** **Cloud SQL read replicas** are dedicated, read-only copies of a primary database instance. Data written to the primary instance is asynchronously replicated to one or more read replicas within the same region or across different regions. ### **How Read Replicas Meet the Requirement** - **Offloads read traffic (✓):** Read-intensive workloads—such as catalog searches, customer lookups, and reporting—can be routed to read replicas instead of the primary instance. - **Eliminates compute contention (✓):** Offloading heavy `SELECT` queries frees up CPU, memory, and I/O resources on the primary instance, allowing critical write transactions to execute without performance degradation. - **Reduces query latency (✓):** Distributing read traffic across multiple replica instances decreases queue times and improves response times for end users. ### **Key Advantages and Capabilities** - **Read scalability:** Scale horizontal read throughput by adding multiple read replicas as application traffic increases. - **Geographic proximity:** Cross-region read replicas can serve read requests closer to global users, reducing network latency. - **Independent compute sizing:** Replicas can be provisioned with different machine types optimized for specific read or reporting patterns. - **Promotion capabilities:** A read replica can be promoted to a standalone primary database instance if needed during regional disaster recovery operations. ### **Why This Is Optimal** While High Availability (HA) configurations protect against infrastructure downtime and automated backups safeguard against data loss, only **read replicas** are purpose-built to scale read throughput and lower query latency under heavy read load.