AWS Certified DevOps Engineer – Professional — Question 140
A company has deployed a new Amazon API Gateway API that retrieves the cost of items for the company's online store. An AWS Lambda function supports the API and retrieves the data from an Amazon DynamoDB table. The API's latency increases during times of peak usage each day. However, the latency of the DynamoDB table reads is constant throughout the day.
A DevOps engineer configures DynamoDB Accelerator (DAX) for the DynamoDB table, and the API latency decreases throughout the day. The DevOps engineer then configures Lambda provisioned concurrency with a limit of two concurrent invocations. This change reduces the latency during normal usage. However, the company is still experiencing higher latency during times of peak usage than during times of normal usage.
Which set of additional steps should the DevOps engineer take to produce the LARGEST decrease in API latency?
Answer options
- A. Increase the read capacity of the DynamoDB table. Use AWS Application Auto Scaling to manage provisioned concurrency for the Lambda function.
- B. Enable caching in API Gateway. Stop using provisioned concurrency for the Lambda function.
- C. Delete the DAX cluster for the DynamoDB table. Use AWS Application Auto Scaling to manage provisioned concurrency for the Lambda function.
- D. Enable caching in API Gateway. Use AWS Application Auto Scaling to manage provisioned concurrency for the Lambda function
Correct answer: D
Explanation
Option D is correct because enabling caching in API Gateway can significantly reduce the number of requests hitting the Lambda function during peak times, while Application Auto Scaling can adjust the provisioned concurrency dynamically to handle increased load. Option A might improve DynamoDB performance but doesn't address the Lambda function's concurrency. Option B would eliminate the benefits of provisioned concurrency, likely worsening latency. Option C removes DAX, which is counterproductive since DAX is meant to improve DynamoDB performance.