MLS-C01 Question 284
Single answerYou have deployed a machine learning model on an Amazon SageMaker endpoint in a production environment. After monitoring the endpoint's performance for several weeks, you notice that the CPU utilization is consistently below 10% and the average throughput is much lower than anticipated. How can you optimize the endpoint to reduce costs while maintaining performance?
- A
Switch the endpoint to use a smaller instance type with fewer compute resources.
- B
Increase the instance type to improve throughput and reduce latency.
- C
Enable Amazon SageMaker Multi-Model endpoints to serve multiple models on the same instance.
- D
Switch the endpoint to use Elastic Inference to accelerate predictions.
Show answer and explanation
Correct answer: A
Explanation
When resource utilization is consistently low, switching to a smaller instance type is a best practice to reduce costs while maintaining adequate performance. Monitoring metrics such as CPU utilization and throughput helps identify opportunities to optimize resources and rightsize your deployment.
- A. Correct.
Switching to a smaller instance type is the correct approach to rightsize resources when utilization is consistently low. This reduces costs while maintaining the necessary level of performance.
- B. Incorrect.
Increasing the instance type would unnecessarily raise costs without addressing the issue of underutilization.
- C. Incorrect.
Multi-Model endpoints are useful for serving multiple models on a single instance but are not directly related to rightsizing an underutilized endpoint.
- D. Incorrect.
Elastic Inference is designed to add GPU acceleration to models using CPU-only instances, which is not relevant to the low utilization issue described.