CloudGuild · Blog · Cheat sheets · Lessons · Certifications
Navigating Machine Learning Deployment with AWS: A SAA-C03 Exam Question Breakdown
Learn how to tackle AWS SAA-C03 questions about machine learning deployment with this detailed breakdown of a sample exam question.
Deploying machine learning models in AWS can be tricky. Candidates often stumble on decisions regarding service combinations, especially when scaling is essential. Let's clarify this with a specific question from the SAA-C03 exam.
The question
A startup is deploying a machine learning model using AWS SageMaker. They require that the model be accessible via a REST API and need to manage the scaling of the inference workload. Which combination of services should they use?
- A. Amazon API Gateway and AWS Lambda
- B. Amazon SageMaker Endpoints and AWS Lambda
- C. Amazon SageMaker Endpoints and Amazon API Gateway
- D. Amazon ECS and Amazon CloudFront
Think before you scroll
Before choosing an option, consider the requirements: accessibility via a REST API and scaling of the inference workload. Which services are designed to work together to meet these needs effectively?
The answer
The correct option is C. Amazon SageMaker Endpoints and Amazon API Gateway. This combination allows for real-time inference through SageMaker Endpoints while the API Gateway manages the REST API and scaling requirements.
Why the other options lose
- A. Amazon API Gateway and AWS Lambda: While API Gateway can expose endpoints, AWS Lambda does not directly provide real-time inference for machine learning models. It’s better suited for serverless functions, not for invoking SageMaker directly.
- B. Amazon SageMaker Endpoints and AWS Lambda: This option pairs SageMaker with Lambda, but Lambda does not manage API exposure. This combination lacks the necessary REST API management that API Gateway offers.
- D. Amazon ECS and Amazon CloudFront: This option is not suitable for machine learning inference. ECS is for container orchestration, and CloudFront is a content delivery network. Neither service is designed for real-time machine learning model inference.
The concept behind it
The key principle here is understanding how AWS services integrate for specific tasks. SageMaker Endpoints are designed for real-time predictions, and API Gateway provides the necessary interface for RESTful access. When designing architectures, always consider how services complement each other based on their primary functions.
Exam trap to remember
Remember: when dealing with machine learning deployments, always look for combinations that ensure real-time inference and API accessibility. The two-question rule applies here: can each service fulfill its role independently while integrating effectively with others?