AWS Certified Machine Learning - Specialty Exam: SageMaker Notebook Instance Management for Cost Optimization and Security

Customizing SageMaker Notebook Instances for Cost Optimization and Security

Question

You work as a machine learning specialist for a security firm that requires you to encrypt all of your machine learning infrastructure in transit and at rest.

Your team is building a fraud detection algorithm using the Random Cut Forest SageMaker built-in algorithm.

You and your teammates are using SageMaker notebook instances to build your model components.

You need to customize the operating system of your notebook instances by installing custom libraries and setting specific operating system level configurations to meet your firm's security requirements.

Your Chief Financial Officer wants to keep the cost of running your SageMaker instances as low as possible.

Therefore, you are required to manage the runtime of your SageMaker notebook instances, only having them running when they are actively in use.

How can you meet your requirements most efficiently?

Answers

Explanations

Click on the arrows to vote for the correct answer

A. B. C. D.

Correct Answer: C.

Option A is incorrect.

When you stop your SageMaker notebook instances, customizations to the operating system, such as installed custom libraries or operating system level settings, are lost.

Option B is incorrect.

Keeping your SageMaker notebook instances running until your team has completed building your Random Cut Forest model will not meet your chief financial officer's requirement of keeping the cost of running your SageMaker instances as low as possible.

Option C is correct.

To avoid losing the installation of custom libraries and setting specific operating system level configurations, you can use a lifecycle configuration to automate customizations of your notebook instances.

Option D is incorrect.

When you terminate your SageMaker notebook instances, the snapshot and the ML storage volume are deleted, thereby deleting your installation of custom libraries and setting specific operating system level configurations.

References:

Please see the Amazon SageMaker developer guide titled Random Cut Forest (RCF) Algorithm (https://docs.aws.amazon.com/sagemaker/latest/dg/randomcutforest.html),

The Amazon SageMaker developer guide titled Notebook instances and SageMaker jobs (https://docs.aws.amazon.com/sagemaker/latest/dg/encryption-at-rest-nbi.html)

The question requires you to customize the operating system of SageMaker notebook instances, encrypt machine learning infrastructure in transit and at rest, and manage the runtime of notebook instances to keep the cost low.

Option A suggests stopping notebook instances at the end of each day and starting them again at the beginning of the next workday to save costs. This option can help reduce costs but it may not be the most efficient approach because you would need to manually perform customizations to the operating system every time you stop and start your notebook instances. Additionally, stopping instances each day can be inconvenient for the team.

Option B suggests keeping SageMaker notebook instances running until your team completes building the Random Cut Forest model. This approach ensures that the team has access to the instances at all times, but it can result in unnecessary costs if the team is not actively using the instances.

Option C suggests using a lifecycle configuration to automate customizations of your notebook instances. This approach allows you to automatically apply customizations to the operating system of your instances, and you can stop instances at the end of each workday and start them again at the beginning of the next workday. This can help reduce costs because instances are only running when they are needed. You can also ensure that your machine learning infrastructure is encrypted in transit and at rest using AWS security services, such as AWS KMS and SSL/TLS.

Option D suggests terminating SageMaker notebook instances at the end of each workday and recreating them at the start of each workday. This approach can also help reduce costs, but it may not be the most efficient because you would need to recreate the notebook instances each day, which can take time and disrupt the team's workflow. Additionally, you would need to ensure that all customizations to the operating system are re-applied each time you recreate the instances.

Therefore, option C is the most efficient approach to meet the requirements of the question because it allows you to automate customizations to the operating system, stop instances when they are not needed, and ensure that machine learning infrastructure is encrypted in transit and at rest.