MLS-C01 Question 291
Select 3You are building a machine learning model to predict customer churn and plan to use Amazon SageMaker for training and deployment. To follow AWS best practices for cost efficiency, security, and scalability, which of the following actions should you take?
- A
Use Spot Instances for training jobs to reduce costs.
- B
Store sensitive training data directly on the local disk of the training instance.
- C
Enable model versioning by utilizing SageMaker Model Registry.
- D
Encrypt data at rest and in transit using AWS Key Management Service (KMS) and SSL/TLS.
- E
Use a single large instance for both training and inference to simplify operations.
Show answer and explanation
Correct answers: A, C, D
Explanation
To follow AWS best practices, you should optimize costs, ensure security, and design for scalability. Using Spot Instances for training jobs helps reduce costs, SageMaker Model Registry ensures scalable model management, and encrypting data protects sensitive information. Storing sensitive data on local disks or combining training and inference on a single instance violates AWS best practices for security and resource management.
- A. Correct.
Spot Instances can significantly reduce costs for training jobs by taking advantage of unused EC2 capacity. This is aligned with AWS cost optimization best practices.
- B. Incorrect.
Storing sensitive training data on the local disk of the training instance is not secure and does not follow AWS security best practices. Data should be stored securely in services like Amazon S3 with proper encryption.
- C. Correct.
SageMaker Model Registry enables you to track model versions, manage metadata, and automate deployment pipelines, adhering to AWS best practices for scalability and maintainability.
- D. Correct.
Encrypting data at rest and in transit ensures that sensitive information is protected, following AWS security best practices.
- E. Incorrect.
Using a single large instance for both training and inference is neither cost-effective nor scalable. AWS recommends separating training and inference workloads to optimize resource usage and scalability.