An API Gateway REST API invokes a Lambda function that loads a large dataset from a DynamoDB table during initialization, causing cold starts of 8–10 seconds. The DynamoDB table uses DAX. The app experiences intermittent long responses and has widely varying traffic (a mid-day spike 10× higher than normal and near-end-of-day drop to 10% of normal). The DevOps engineer must reduce Lambda latency at all times of day. Which solution meets the requirement?

The correct solution is to enable provisioned concurrency on the Lambda function and configure Application Auto Scaling for it. Provisioned concurrency keeps Lambda functions initialized and ready to respond, eliminating cold starts. Auto Scaling ensures that the number of pre-initialized instances dynamically adjusts to traffic patterns (spikes and drops), maintaining low latency during varying load. Incorrect options: Enabling provisioned concurrency with a value of 1 would only help a single invocation and wouldn't scale with traffic. Deleting the DAX cluster would likely increase DynamoDB access latency, worsening the problem. Setting reserved concurrency to 0 would prevent the function from running at all, which is incorrect. Setting reserved concurrency on the Lambda function limits its maximum concurrent executions but doesn't address cold starts. Configuring Application Auto Scaling on API Gateway manages API Gateway capacity, not Lambda cold starts.