MLA-C01 Question 424
Select 3You are working as a Machine Learning Engineer at a company that has deployed multiple ML models on Amazon SageMaker. The team needs to monitor and allocate costs for different projects and departments effectively. Which of the following techniques can help you achieve this goal?
- A
Apply resource tags to SageMaker instances and use AWS Cost Explorer to filter costs by tags.
- B
Enable AWS Cost and Usage Reports for detailed billing data, including resource-level information.
- C
Deploy models in separate AWS accounts for each department and enable consolidated billing.
- D
Use AWS Trusted Advisor to automatically allocate costs to specific projects.
- E
Create a custom CloudWatch dashboard to monitor costs by SageMaker instance type.
Show answer and explanation
Correct answers: A, B, C
Explanation
To effectively track and allocate costs for ML workloads on SageMaker, you can use resource tagging combined with AWS Cost Explorer, enable detailed AWS Cost and Usage Reports, or separate projects by accounts and use consolidated billing for aggregation. AWS Trusted Advisor and CloudWatch are not suitable tools for cost tracking or allocation in this scenario.
- A. Correct.
Correct. Resource tagging allows you to categorize SageMaker resources, such as instances, by project or department. AWS Cost Explorer can then filter and allocate costs based on these tags.
- B. Correct.
Correct. AWS Cost and Usage Reports provide detailed cost data, including resource-level usage, which can help in tracking and allocating costs for ML projects deployed on SageMaker.
- C. Correct.
Correct. Deploying models in separate AWS accounts for departments allows for clear cost segregation. Consolidated billing can then aggregate and report costs across accounts.
- D. Incorrect.
Incorrect. AWS Trusted Advisor provides recommendations for cost optimization, security, and performance but does not allocate costs to specific projects or departments.
- E. Incorrect.
Incorrect. While CloudWatch can monitor instance metrics like CPU and memory usage, it is not designed for cost tracking or allocation purposes.