MLA-C01 Question 237
Single answerYou are deploying a machine learning model to production using Amazon SageMaker. The model is expected to serve thousands of real-time predictions per second with minimal latency. Additionally, you need the ability to perform A/B testing between two model versions to evaluate performance. Which of the following approaches would best meet these requirements?
- A
Deploy the models as separate SageMaker batch transform jobs and analyze the results.
- B
Deploy both model versions to a single endpoint using SageMaker multi-model endpoints and configure traffic splitting.
- C
Deploy each model version to separate SageMaker endpoints and use Amazon Route 53 to split traffic between them.
- D
Deploy both model versions to a SageMaker endpoint using production variants and configure traffic distribution.
Show answer and explanation
Correct answer: D
Explanation
Amazon SageMaker production variants are specifically designed for scenarios where you need to deploy multiple model versions to a single endpoint and split traffic between them for A/B testing. This approach minimizes latency and simplifies the deployment process compared to other options like batch transform jobs or using separate endpoints with Amazon Route 53.
- A. Incorrect.
Batch transform jobs are designed for offline inference and are not suitable for real-time predictions, making this approach infeasible for the given requirements.
- B. Incorrect.
SageMaker multi-model endpoints are used to serve multiple models from a single endpoint, but they are not specifically designed for A/B testing or real-time traffic splitting.
- C. Incorrect.
While using separate endpoints with Amazon Route 53 for traffic splitting is technically possible, it adds unnecessary complexity and latency compared to SageMaker’s built-in production variants.
- D. Correct.
Using SageMaker production variants allows you to deploy multiple model versions to the same endpoint and easily configure traffic distribution for A/B testing, meeting both the real-time prediction and A/B testing requirements.