Google Professional Machine Learning Engineer Question 340
Select 2Google Cloud PlatformYou are deploying a machine learning model using Vertex AI Prediction to serve real-time predictions. Your application experiences fluctuating traffic patterns, with peaks during specific hours of the day. To ensure optimal performance and cost efficiency, how can you configure the serving backend to handle the varying throughput?
- A
Configure auto-scaling for the deployed model and set minimum and maximum replica limits.
- B
Manually monitor the traffic and adjust the number of replicas as needed.
- C
Enable the 'Always-on' setting to ensure a fixed number of replicas are always running.
- D
Use Vertex AI’s auto-scaling feature to automatically adjust the number of replicas based on traffic load.
- E
Deploy the model to a single-node Kubernetes cluster to simplify scaling.
Show answer and explanation
Correct answers: A, D
Explanation
To handle fluctuating traffic patterns effectively with Vertex AI Prediction, it is crucial to leverage its auto-scaling capabilities. Configuring auto-scaling with minimum and maximum replica limits allows the platform to dynamically adjust resources, ensuring optimal performance during peak traffic while avoiding unnecessary costs during low traffic. Manual adjustments or fixed replica settings are not efficient or practical solutions for dynamic workloads.
- A. Correct.
Correct. Configuring auto-scaling with defined minimum and maximum replica limits allows the system to automatically adjust the number of replicas based on traffic patterns, ensuring cost efficiency and performance.
- B. Incorrect.
Incorrect. Manually adjusting the number of replicas is labor-intensive and error-prone, especially with fluctuating traffic patterns. It is not a recommended practice for scalable systems.
- C. Incorrect.
Incorrect. The 'Always-on' setting ensures a fixed number of replicas but does not handle fluctuating traffic dynamically, leading to potential over-provisioning during low traffic or under-provisioning during high traffic.
- D. Correct.
Correct. Vertex AI’s auto-scaling feature is specifically designed to dynamically manage the number of replicas based on traffic load, making it ideal for handling fluctuating throughput.
- E. Incorrect.
Incorrect. Deploying the model to a single-node Kubernetes cluster does not provide the necessary scalability to handle fluctuating traffic patterns effectively.