MLS-C01 Question 248
Single answerYou are managing a production machine learning model hosted on Amazon SageMaker. The model is experiencing degraded performance in providing predictions due to a significant increase in traffic. You need to ensure the system can handle the traffic while minimizing downtime and avoiding over-provisioning of resources. What is the most appropriate solution?
- A
Manually provision additional instances for the endpoint in Amazon SageMaker.
- B
Enable SageMaker endpoint auto-scaling to automatically adjust the number of instances based on traffic.
- C
Use Amazon CloudWatch to monitor traffic and manually scale the endpoint when needed.
- D
Deploy the model to an Amazon EC2 instance and use an application load balancer to handle the traffic.
Show answer and explanation
Correct answer: B
Explanation
Amazon SageMaker endpoint auto-scaling is designed to handle fluctuating traffic automatically by adjusting the number of instances based on metrics like CPU or memory utilization. This approach eliminates manual intervention, reduces downtime, and avoids over-provisioning, making it the most suitable solution for managing production machine learning model deployments in SageMaker.
- A. Incorrect.
Manually provisioning additional instances requires constant monitoring and intervention, which is not efficient, especially for traffic patterns that fluctuate dynamically.
- B. Correct.
Enabling SageMaker endpoint auto-scaling is the most appropriate solution because it allows automatic adjustment of resources based on predefined metrics, such as CPU or memory utilization, minimizing downtime and avoiding resource over-provisioning.
- C. Incorrect.
Using Amazon CloudWatch for monitoring and manually scaling still requires human intervention, which can lead to delays in responding to sudden traffic changes.
- D. Incorrect.
Deploying the model to an Amazon EC2 instance with a load balancer is possible, but it requires significant operational overhead and moves away from the managed services provided by SageMaker.