MLS-C01 Question 252
Select 2You are designing a machine learning solution to predict customer churn for a global e-commerce platform. The solution must ensure high availability, scalability to handle millions of requests per day, and fault tolerance to minimize downtime. You have decided to deploy the model using Amazon SageMaker. Which combination of configurations will best achieve these requirements?
- A
Deploy the model on an Amazon SageMaker endpoint with Multi-Model Endpoint support.
- B
Use Amazon SageMaker Automatic Scaling to adjust the number of instances based on incoming traffic.
- C
Enable Amazon SageMaker endpoint variants and distribute traffic across multiple variants using traffic shifting.
- D
Deploy the model on an Amazon EC2 instance to have greater control over instance configuration.
- E
Use AWS Elastic Load Balancer (ELB) in front of the Amazon SageMaker endpoint to distribute traffic.
Show answer and explanation
Correct answers: B, C
Explanation
To build an ML solution that is performant, scalable, and fault-tolerant, Amazon SageMaker Automatic Scaling ensures that the endpoint can handle changes in traffic dynamically. Endpoint variants with traffic shifting provide an additional layer of fault tolerance by enabling traffic redirection during failures or updates. These configurations align closely with the requirements of high availability and scalability, making them the best choices for this scenario.
- A. Incorrect.
Multi-Model Endpoints are useful for hosting multiple models on a single endpoint, but they are not specifically designed to enhance fault tolerance or scalability for high-traffic scenarios.
- B. Correct.
Amazon SageMaker Automatic Scaling ensures that the endpoint can dynamically adjust resources to handle traffic spikes, improving scalability and availability.
- C. Correct.
Using endpoint variants and traffic shifting in Amazon SageMaker allows for fault tolerance and A/B testing by redirecting traffic between multiple variants as needed.
- D. Incorrect.
While deploying on Amazon EC2 might provide more control, it requires significant manual effort to ensure scalability, availability, and fault tolerance, which makes it less suitable than SageMaker for this scenario.
- E. Incorrect.
AWS Elastic Load Balancer cannot be directly used with Amazon SageMaker endpoints. SageMaker endpoints are already managed services that include load balancing capabilities.