MLA-C01 Question 235
Select 2You are deploying a machine learning model for real-time inference using Amazon SageMaker. The model requires high availability and needs to automatically scale based on incoming requests. Additionally, you need to monitor endpoint performance and model metrics (e.g., latency, invocation count) to ensure smooth operation. Which combination of actions should you take to meet these requirements?
- A
Deploy the model to an Amazon SageMaker endpoint and configure auto-scaling policies based on metrics like CPU utilization or invocation count.
- B
Use Amazon CloudWatch to monitor endpoint performance metrics such as latency and invocation count.
- C
Enable Amazon SageMaker Model Monitor to capture and analyze model drift and input data quality.
- D
Use AWS Lambda to handle auto-scaling for your SageMaker endpoint.
- E
Configure an Elastic Load Balancer (ELB) in front of the SageMaker endpoint for high availability.
Show answer and explanation
Correct answers: A, B
Explanation
To deploy a machine learning model with high availability and auto-scaling capabilities, Amazon SageMaker endpoints should be used with auto-scaling policies configured based on metrics like CPU utilization and invocation count. Monitoring endpoint performance can be achieved through Amazon CloudWatch, which provides detailed metrics for latency, invocation count, and error rates. Other options, such as SageMaker Model Monitor and AWS Lambda, address different use cases and are not applicable here.
- A. Correct.
Correct. Amazon SageMaker endpoints support auto-scaling policies that can be configured based on custom metrics like CPU utilization or invocation count, ensuring scalability based on demand.
- B. Correct.
Correct. Amazon CloudWatch is the recommended tool for monitoring SageMaker endpoint performance metrics such as latency, invocation count, and error rates.
- C. Incorrect.
Incorrect. While SageMaker Model Monitor is useful for monitoring data drift and model quality, it is not directly related to endpoint performance or auto-scaling.
- D. Incorrect.
Incorrect. AWS Lambda is not used for auto-scaling SageMaker endpoints. SageMaker endpoints natively support auto-scaling through pre-configured policies.
- E. Incorrect.
Incorrect. Elastic Load Balancers are not used with SageMaker endpoints because SageMaker endpoints are already managed services that ensure high availability.