An application uses an Aurora MySQL cluster with a single read replica. Read performance worsens when connections exceed 200; normal load is about 180 connections but sometimes spikes above 200. The administrator needs the system to scale automatically as user connections rise and fall. Which approach satisfies this requirement?

Creating an auto scaling policy that targets a DatabaseConnections metric value of 195 is the most effective solution. This allows Aurora to automatically add or remove read replicas based on connection load, ensuring performance is maintained during spikes and optimizing costs during normal operation. Increasing the instance size of the existing replica might temporarily alleviate the issue but won't scale dynamically and could be over-provisioned. Switching to serverless mode when connections exceed 200 isn't a standard or practical way to handle read replica scaling for an existing provisioned cluster. Moving to multiple writer instances is for write scalability, not read, and Aurora MySQL only supports a single writer.