MLA-C01 Question 260
Select 2You are a machine learning engineer working for a retail company that needs to serve a product recommendation model. The model will provide real-time recommendations on the website and also generate a daily batch of recommendations for email campaigns. Which combination of services and methods would best meet these requirements?
- A
Use Amazon SageMaker Hosting Services to deploy the model for real-time inference and AWS Glue for batch processing.
- B
Use AWS Lambda with an API Gateway for real-time inference and Amazon SageMaker Batch Transform for batch processing.
- C
Use Amazon SageMaker Multi-Model Endpoint for real-time inference and Amazon EMR for batch processing.
- D
Use Amazon ECS with a Flask application for real-time inference and Amazon Athena for batch processing.
- E
Use Amazon SageMaker Hosting Services for real-time inference and Amazon SageMaker Processing Jobs for batch processing.
Show answer and explanation
Correct answers: A, B
Explanation
To serve a machine learning model in both real-time and batch scenarios, you need to select services that are optimized for these specific use cases. Amazon SageMaker Hosting Services and AWS Lambda with API Gateway are ideal for real-time inference, each offering scalability and low-latency responses. For batch processing, AWS Glue and Amazon SageMaker Batch Transform are highly efficient options designed for handling large datasets and offline inference. Combining these services ensures a reliable and cost-effective solution for both real-time and batch inference needs.
- A. Correct.
Amazon SageMaker Hosting Services is well-suited for real-time inference as it provides scalable endpoints, while AWS Glue is an appropriate service for batch processing large datasets using ETL workflows.
- B. Correct.
AWS Lambda with an API Gateway is a lightweight and cost-effective way to handle real-time inference, and Amazon SageMaker Batch Transform is specifically designed for large-scale, offline batch processing.
- C. Incorrect.
Amazon SageMaker Multi-Model Endpoint is used for serving multiple models on the same endpoint for real-time inference, but it is not specifically designed for batch processing. Amazon EMR is a big data processing service and may not be the most efficient choice for typical ML batch inference workloads.
- D. Incorrect.
Amazon ECS with a Flask application can handle real-time inference, but setting up and managing this infrastructure requires more effort compared to managed services. Amazon Athena is a query service and is not typically used for batch inference.
- E. Incorrect.
Amazon SageMaker Hosting Services is suitable for real-time inference, but SageMaker Processing Jobs are primarily designed for data preprocessing, model evaluation, or data transformation, not for serving batch inference.