MLA-C01 Question 241
Single answerA company has trained a machine learning model for real-time fraud detection and needs to deploy it with low latency requirements. The current architecture includes an application hosted on Amazon ECS and a high volume of API requests. Which deployment infrastructure is the most suitable given the requirements?
- A
Deploy the model using Amazon SageMaker Batch Transform.
- B
Host the model as an endpoint using Amazon SageMaker Hosting Services.
- C
Deploy the model as a container on Amazon ECS with an attached Elastic Load Balancer.
- D
Deploy the model on an Amazon EC2 instance with an attached API Gateway.
Show answer and explanation
Correct answer: C
Explanation
Since the company’s architecture already uses Amazon ECS, deploying the model as a container on ECS with an Elastic Load Balancer is the most suitable option. It ensures low latency, handles high API request volumes, and integrates seamlessly into the existing infrastructure without requiring major changes.
- A. Incorrect.
Amazon SageMaker Batch Transform is designed for batch processing and not suitable for real-time, low-latency requirements.
- B. Incorrect.
Amazon SageMaker Hosting Services can provide low-latency endpoints, but since the architecture already uses Amazon ECS, this would require significant architectural changes.
- C. Correct.
Deploying the model as a container on Amazon ECS with an attached Elastic Load Balancer aligns with the existing architecture, supports high API request volumes, and meets low-latency requirements.
- D. Incorrect.
Deploying on Amazon EC2 with an API Gateway is possible but would require more manual setup and management, and it might not align well with the existing container-based ECS architecture.