Google Professional Machine Learning Engineer Question 345
Select 3Google Cloud PlatformYou are deploying a machine learning model using a containerized serving approach on Google Cloud. The model needs to serve predictions at scale with low latency, handle traffic spikes, and integrate seamlessly with other Google Cloud services. Which of the following steps should you take to achieve this deployment?
- A
Use Google Kubernetes Engine (GKE) to orchestrate the containerized model serving.
- B
Deploy the containerized model directly to a Compute Engine VM without any orchestration.
- C
Use Cloud Run to deploy the containerized model for fully managed serverless scaling.
- D
Integrate the deployment with Cloud Monitoring to track performance and latency.
- E
Manually scale the serving infrastructure by pre-allocating resources for peak traffic.
Show answer and explanation
Correct answers: A, C, D
Explanation
To deploy a containerized model serving solution at scale on Google Cloud, you should use orchestration tools like GKE for flexible deployments or Cloud Run for fully managed serverless scaling. These solutions provide automated scaling and integration with monitoring tools like Cloud Monitoring, ensuring performance and reliability. Manually scaling or deploying directly to a VM is inefficient and does not align with Google Cloud best practices for scalable ML deployments.
- A. Correct.
Using Google Kubernetes Engine (GKE) is a suitable option for orchestrating containerized workloads, providing horizontal scaling and integration with other Google Cloud services such as Cloud Monitoring.
- B. Incorrect.
Deploying directly to a Compute Engine VM without orchestration is not recommended for scalable and automated deployments, as it lacks the benefits of managed scaling and efficiency provided by platforms like GKE or Cloud Run.
- C. Correct.
Cloud Run is an ideal option for containerized model serving if you need a fully managed, serverless solution that automatically scales with traffic while maintaining low latency.
- D. Correct.
Integrating with Cloud Monitoring is critical for tracking the performance, availability, and latency of your model serving deployment, ensuring reliability at scale.
- E. Incorrect.
Manually scaling infrastructure is not efficient or recommended for handling traffic spikes, as it does not leverage the automated scaling capabilities provided by GKE or Cloud Run.