An ecommerce application uses an Amazon Aurora cluster with memory-optimized instances (one writer and one reader). During sudden traffic spikes CloudWatch shows high RAM usage and increased SELECT latency. The change must improve DB cluster performance with minimal downtime and without data loss. Which modification is the best choice?

Adding an Aurora Replica offloads read traffic from the primary instance, directly addressing the high RAM usage and increased SELECT latency. This improves performance without downtime or data loss because replicas are asynchronously updated and can be added to an existing cluster. Converting to a multi-master cluster is a more complex change, primarily for high availability and write scaling, and might involve downtime. Creating a snapshot and restoring to larger instances would involve significant downtime during the restoration process. Doubling disk storage capacity is irrelevant to high RAM usage and SELECT latency, as these issues are related to compute and memory, not storage space.