MLA-C01 Practice Question 68: Airflow Orchestration Choice
AWS Certified Machine Learning Engineer – Associate (MLA-C01) · Test 7 · question 8 of 10
A company's data engineering team runs about 200 Apache Airflow DAGs on Amazon Managed Workflows for Apache Airflow (Amazon MWAA). One nightly DAG builds the feature tables for a demand model. The ML team now wants that same DAG to start a SageMaker AI training job after the features are built, then run a batch transform job with the new model. The company has a standing rule against adding a second workflow orchestration tool.
How should the ML team add the steps?
Show answer and explanation
Answer:
The Apache Airflow community provides operators, plugins that simplify connecting to services, that integrate Airflow with AWS services including Amazon SageMaker AI, and MWAA supports them. Adding training and batch transform tasks to the existing DAG keeps the ML steps in the orchestrator the company already runs. The steps depend directly on the feature-building tasks, share the DAG's retries and monitoring, and add no new tool (option C).
- A is wrong because SageMaker Pipelines is a second workflow orchestration tool, which the company's rule forbids, and the dependency between the two workflows now runs through an event. - B is wrong for the same reason. Step Functions is another orchestrator, and S3 events fire per object, which is a fragile signal that a multi-table build has finished. - D is wrong because a cron job guesses when the DAG finishes instead of depending on it. When the feature build runs late or fails, training runs on stale or partial data, and nobody is alerted.
Reference: What Is Amazon Managed Workflows for Apache Airflow? - Integration