MLS-C01 Question 255
Select 2You are a Machine Learning Engineer managing an ML model hosted on Amazon SageMaker. The model's endpoint has recently experienced an increase in latency and errors. You want to monitor the endpoint’s performance and identify the root cause in near real-time. Which combination of actions will help you achieve this?
- A
Enable Amazon CloudWatch Metrics for the SageMaker endpoint to monitor latency and error rates.
- B
Set up AWS CloudTrail to log API calls made to the SageMaker endpoint.
- C
Use Amazon SageMaker Model Monitor to capture and analyze data drift in the endpoint traffic.
- D
Enable detailed monitoring in Amazon CloudWatch for additional metrics on endpoint performance.
- E
Configure an Amazon SNS topic to send notifications when a specific latency threshold is breached.
Show answer and explanation
Correct answers: A, C
Explanation
Monitoring and identifying performance issues in a SageMaker endpoint requires leveraging tools that provide real-time metrics and insights. Amazon CloudWatch Metrics allows you to track latency and error rates, while SageMaker Model Monitor can analyze data drift or other anomalies affecting the endpoint. These tools together provide a comprehensive approach to diagnosing and resolving performance issues.
- A. Correct.
Enabling Amazon CloudWatch Metrics will allow you to monitor metrics such as latency and error rates for your SageMaker endpoint. This is critical for understanding the performance issues.
- B. Incorrect.
AWS CloudTrail logs API activity, but it does not provide performance metrics such as latency or error rates, making it insufficient for this specific use case.
- C. Correct.
Amazon SageMaker Model Monitor is designed to capture and analyze aspects like data drift or other anomalies in real-time. It can help identify potential issues contributing to endpoint performance problems.
- D. Incorrect.
Enabling detailed monitoring in CloudWatch is not applicable to SageMaker endpoints as they already provide specific metrics by default through CloudWatch.
- E. Incorrect.
Configuring an Amazon SNS topic for alerts is useful for notifications, but it does not directly help in monitoring or identifying the root cause of the latency and errors.