MLS-C01 Question 268
Select 2Your company has trained an advanced machine learning model for fraud detection and needs to deploy it globally with low latency for end-users in multiple regions. The model should be highly available and resilient to failures in any Availability Zone (AZ) or even an entire AWS Region. Which combination of solutions would best meet these requirements?
- A
Deploy the model to multiple AWS Regions using Amazon SageMaker endpoints and configure Amazon Route 53 for latency-based routing.
- B
Use Amazon SageMaker Multi-Model Endpoints to deploy all regional models in a single AWS Region and rely on cross-region traffic routing.
- C
Deploy the model to multiple Availability Zones within a single AWS Region using Amazon SageMaker endpoints and an Application Load Balancer.
- D
Set up the model in multiple AWS Regions using Amazon SageMaker endpoints and configure Amazon CloudFront with regional edge caches for latency optimization.
- E
Run the model in a single AWS Region and use AWS Global Accelerator to route traffic from users in other regions to the deployment.
Show answer and explanation
Correct answers: A, D
Explanation
To ensure low latency for global users and high availability for the fraud detection model, deploying to multiple AWS Regions is essential. Amazon SageMaker endpoints in multiple regions combined with Amazon Route 53 or Amazon CloudFront for optimized routing provide a robust, resilient, and scalable solution. This setup ensures requests are served from the nearest region, minimizing latency and mitigating the impact of regional or AZ-level failures.
- A. Correct.
Correct. Deploying the model to multiple AWS Regions ensures low latency for global users, and Amazon Route 53's latency-based routing allows requests to be sent to the nearest region, providing high availability even if an entire region goes down.
- B. Incorrect.
Incorrect. While Amazon SageMaker Multi-Model Endpoints can host multiple models, placing all deployments in a single region results in increased latency for users in geographically distant regions and does not provide regional fault tolerance.
- C. Incorrect.
Incorrect. Deploying to multiple AZs in a single region improves availability within that region but does not address global latency or regional failures.
- D. Correct.
Correct. Deploying to multiple AWS Regions and using Amazon CloudFront with regional edge caches enhances latency for global users and maintains high availability by routing requests to the nearest region with minimal delays.
- E. Incorrect.
Incorrect. AWS Global Accelerator can improve networking performance but does not solve latency issues for global users when the model is deployed in a single region. It also doesn't provide fault tolerance across regions.