MLS-C01 Question 269
Select 3You are deploying a machine learning model using Amazon SageMaker to serve predictions for users across multiple AWS Regions and ensure high availability in each region. Which combination of strategies should you use to achieve this goal?
- A
Deploy the model endpoint in each AWS Region and use Amazon Route 53 to route traffic based on latency.
- B
Use AWS CloudFormation StackSets to deploy the infrastructure, including SageMaker endpoints, across multiple AWS Regions.
- C
Deploy the model endpoint in a single AWS Region and use AWS Global Accelerator to route traffic globally.
- D
Configure SageMaker Multi-Model Endpoints to balance traffic across multiple AWS Regions.
- E
Enable Auto Scaling for SageMaker endpoints within each AWS Region to handle variable traffic loads.
Show answer and explanation
Correct answers: A, B, E
Explanation
To serve predictions across multiple AWS Regions and ensure high availability, you need to deploy SageMaker endpoints in each region where your users are located. Using Amazon Route 53 for latency-based routing ensures users are directed to the nearest endpoint. AWS CloudFormation StackSets simplifies the deployment process across multiple regions, while enabling Auto Scaling ensures that each endpoint can scale to meet traffic demands dynamically. These strategies collectively address both low latency and high availability.
- A. Correct.
Correct: Deploying the model endpoint in each AWS Region ensures that users in different regions can access the endpoints with low latency. Using Amazon Route 53 to route traffic based on latency improves user experience by directing requests to the nearest region.
- B. Correct.
Correct: AWS CloudFormation StackSets can be used to automate and standardize the deployment of SageMaker endpoints and other infrastructure across multiple AWS Regions.
- C. Incorrect.
Incorrect: While AWS Global Accelerator can route traffic globally, deploying the endpoint in only one region would not ensure low latency or high availability for users in other regions.
- D. Incorrect.
Incorrect: SageMaker Multi-Model Endpoints are designed to host multiple models on the same endpoint within a single region. They do not provide traffic balancing across multiple AWS Regions.
- E. Correct.
Correct: Enabling Auto Scaling ensures that SageMaker endpoints within each region can handle variable traffic loads, improving availability and cost efficiency.