MLA-C01 Question 284
Select 3You are tasked with setting up the infrastructure for a machine learning model deployment using Amazon SageMaker. The architecture requires the following: the model must be hosted on an endpoint with auto-scaling enabled, real-time predictions are needed, and the infrastructure must minimize operational overhead. Which steps should you take to create and script the infrastructure based on these requirements?
- A
Use the AWS SDK or AWS CLI to create a SageMaker endpoint and specify the desired instance type for hosting.
- B
Configure an Amazon EC2 instance with an Auto Scaling group and manually deploy the model on the instance.
- C
Enable automatic scaling for the SageMaker endpoint by creating an Application Auto Scaling policy.
- D
Configure a SageMaker endpoint to use model containers in Amazon ECS for real-time predictions.
- E
Write a script using AWS SDK or AWS CloudFormation to automate infrastructure creation and configuration.
Show answer and explanation
Correct answers: A, C, E
Explanation
To meet the requirements of hosting a model with auto-scaling, real-time predictions, and minimal operational overhead, Amazon SageMaker endpoints are the ideal solution. AWS SDK, CLI, or CloudFormation can be used to script the infrastructure, and Application Auto Scaling can be configured to meet the auto-scaling requirement. EC2 with manual deployment and Amazon ECS are less suitable due to higher operational overhead and complexity.
- A. Correct.
Correct: Using the AWS SDK or CLI allows you to create a SageMaker endpoint and specify the desired instance type, which is a requirement for hosting the model.
- B. Incorrect.
Incorrect: EC2 instances with Auto Scaling require more manual configuration and operational overhead, which does not align with the requirement to minimize operational overhead.
- C. Correct.
Correct: Enabling automatic scaling for the SageMaker endpoint via Application Auto Scaling ensures the endpoint can adjust resources based on demand, fulfilling the auto-scaling requirement.
- D. Incorrect.
Incorrect: While Amazon ECS can host model containers, it is not the primary service for hosting SageMaker models for real-time predictions, and it requires additional setup.
- E. Correct.
Correct: Writing a script using AWS SDK or CloudFormation allows you to automate the infrastructure creation and configuration, minimizing operational overhead as required.