Google Professional Machine Learning Engineer Question 375
Select 4Google Cloud PlatformYou are tasked with hosting a third-party MLFlow pipeline on Google Cloud for your company's machine learning workflows. The pipeline needs to track experiments, manage models, and ensure scalability. Which of the following steps are required to successfully deploy and integrate MLFlow on Google Cloud?
- A
Deploy MLFlow on a Google Kubernetes Engine (GKE) cluster to handle scalable workloads.
- B
Use Cloud Storage as the artifact store for experiment tracking and model storage.
- C
Set up a Cloud Function to directly execute MLFlow pipelines.
- D
Configure a Cloud SQL instance as the backend store for MLFlow metadata.
- E
Deploy MLFlow using Vertex AI Pipelines for built-in integration with Google Cloud.
- F
Secure the deployment using Identity and Access Management (IAM) roles for access control.
Show answer and explanation
Correct answers: A, B, D, F
Explanation
To successfully host and integrate a third-party MLFlow pipeline on Google Cloud, you need a scalable deployment platform like GKE, a suitable artifact store like Cloud Storage, and a backend store for metadata like Cloud SQL. Additionally, securing the deployment with IAM roles ensures proper access control. While Vertex AI Pipelines and Cloud Functions are valuable Google Cloud services, they are not directly compatible with hosting MLFlow in this scenario.
- A. Correct.
Deploying MLFlow on Google Kubernetes Engine (GKE) provides the scalability and flexibility needed to handle workloads, making it an appropriate choice for hosting MLFlow on Google Cloud.
- B. Correct.
Cloud Storage can serve as an artifact store for MLFlow, enabling it to store experiment data and model artifacts in a scalable and cost-effective way.
- C. Incorrect.
Cloud Functions are not suitable for hosting or executing MLFlow pipelines because they are designed for event-driven serverless workloads and cannot handle the stateful, long-running operations of MLFlow.
- D. Correct.
Cloud SQL can be used as the backend store for MLFlow metadata, ensuring reliable storage and retrieval of experiment information.
- E. Incorrect.
Vertex AI Pipelines is a managed solution for orchestrating ML workflows, but it is not designed to directly host MLFlow, which requires its own deployment and configuration.
- F. Correct.
Using IAM roles for access control secures the MLFlow deployment and ensures that only authorized users or services can access the pipeline and associated resources.