MLA-C01 Question 349
Single answerYou are deploying a machine learning model using Amazon SageMaker and want to ensure minimal downtime while also having the ability to quickly roll back in case of issues. The model is critical to business operations, and you want to test the new version with a small subset of traffic before gradually shifting all traffic to it. Which deployment strategy should you use?
- A
Blue/Green Deployment
- B
Canary Deployment
- C
Linear Deployment
- D
Rolling Deployment
Show answer and explanation
Correct answer: B
Explanation
Canary Deployment is the most appropriate strategy for this scenario because it provides the ability to test the new model version with a small subset of traffic initially. If the new version performs as expected, traffic can be gradually shifted to it, ensuring minimal risk of downtime or negative impact. Additionally, it allows for easy rollback if any issues are detected during the testing phase.
- A. Incorrect.
Blue/Green Deployment involves creating a completely separate environment for the new version and swapping traffic to it all at once. While it allows rollback, it does not gradually shift traffic and therefore is not suited for testing with a small subset of traffic.
- B. Correct.
Canary Deployment is the correct option because it allows you to route a small percentage of traffic to the new model version initially, test its behavior, and then gradually shift more traffic, ensuring minimal downtime and easier rollback if issues occur.
- C. Incorrect.
Linear Deployment shifts traffic in equal increments over a defined period but does not allow the initial testing with a small subset of traffic that Canary Deployment does.
- D. Incorrect.
Rolling Deployment is not a supported strategy in Amazon SageMaker. It is typically used in other systems for updating software across a fleet of servers incrementally, and it is not applicable to the scenario described.