MLS-C01 Question 267
Select 2You are deploying a machine learning model using Amazon SageMaker and want to ensure high availability and low latency for users across multiple AWS Regions and Availability Zones. Which of the following approaches will meet these requirements?
- A
Deploy the SageMaker endpoint in multiple Regions and use Amazon Route 53 latency-based routing to direct traffic.
- B
Deploy the SageMaker endpoint in a single Region with multiple Availability Zones and use an Application Load Balancer (ALB) to distribute traffic.
- C
Deploy the SageMaker endpoint in multiple Regions and use AWS Global Accelerator to route traffic based on user proximity.
- D
Deploy the SageMaker endpoint in a single Availability Zone and use Amazon CloudFront to cache responses globally.
- E
Configure Amazon SageMaker Multi-Model Endpoints to automatically replicate the model across multiple Regions.
Show answer and explanation
Correct answers: A, C
Explanation
To ensure high availability and low latency for users globally, SageMaker endpoints should be deployed in multiple Regions. Amazon Route 53 latency-based routing and AWS Global Accelerator are both valid solutions for directing users to the nearest endpoint based on latency or proximity. Deploying in a single Region or AZ does not meet the multi-Region requirement, and CloudFront or Multi-Model Endpoints are not designed for this use case.
- A. Correct.
This is correct. Deploying the SageMaker endpoint in multiple Regions and using Route 53's latency-based routing ensures users are directed to the nearest Region for low latency and high availability.
- B. Incorrect.
This is incorrect. While deploying in multiple Availability Zones improves fault tolerance within a single Region, it does not address the need for multi-Region deployments or low latency for users in different geographical locations.
- C. Correct.
This is correct. Deploying SageMaker endpoints in multiple Regions and using AWS Global Accelerator ensures traffic is routed to the closest Region based on proximity, improving latency and availability.
- D. Incorrect.
This is incorrect. Deploying in a single Availability Zone creates a single point of failure, and Amazon CloudFront is not suitable for caching dynamic responses from SageMaker endpoints.
- E. Incorrect.
This is incorrect. SageMaker Multi-Model Endpoints are used to host multiple models on a single endpoint within a Region. They do not automatically replicate models across multiple Regions.