A company runs a microservice as an AWS Lambda function. The microservice writes to an on-premises SQL database that supports only a limited number of concurrent connections. When Lambda invocation volume increases, the database is overwhelmed and crashes, causing downtime. The company has an AWS Direct Connect link between the VPC and the on-premises datacenter. The company wants to protect the database from being overwhelmed. Which solution satisfies these requirements?

The correct solution uses an SQS queue to decouple the Lambda invocation from the database write operation, acting as a buffer. This prevents the database from being overwhelmed during peak Lambda invocation volume. A separate Lambda function can then process messages from the SQS queue at a controlled rate. Configuring a reserved concurrency limit on this processing Lambda function, below the database's maximum supported concurrent connections, ensures that the database is not overloaded. Incorrect options: Migrating to Aurora Serverless is a significant architectural change and doesn't directly address protecting the existing on-premises database. RDS Proxy is for Amazon RDS databases, not on-premises SQL databases. Using SNS to invoke Lambda still directly exposes the database to potential overload if the Lambda concurrency isn't properly managed and throttled, and provisioned concurrency alone doesn't guarantee the database won't be overwhelmed if the provisioned limit is too high or if the database itself is the bottleneck.