Answer-first summary for fast verification
Answer: Deploy the new model version on the same endpoint as the old version, Update the Traffic split percentage to allocate a small portion of traffic to the new version
The correct approach involves deploying your model to an existing endpoint (Option B) and adjusting the Traffic split percentage (Option C) to ensure the new version handles only a small fraction of the traffic. This method minimizes operational overhead and ensures a seamless transition without downtime. Option A is incorrect as creating a Docker container image is unnecessary with AutoML. Option D is not applicable here since Canary Deployment with Cloud Build is a CI/CD pipeline process, which isn't required in this managed environment. Option E would increase operational complexity and is not necessary for achieving the goal. For more details, visit: [Google Cloud Vertex AI Documentation](https://cloud.google.com/vertex-ai/docs/predictions/deploy-model-console).
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
As a developer and data scientist at a medium-sized company utilizing Vertex AI / AI Platform, you've updated an Auto ML model and wish to deploy it to production while keeping both the old and new versions active simultaneously. The new version should handle only a small fraction of the traffic. The solution must minimize operational overhead and ensure seamless transition between model versions without downtime. Which two actions should you take? (Choose two)
A
Save the model in a Docker container image for manual deployment
B
Deploy the new model version on the same endpoint as the old version
C
Update the Traffic split percentage to allocate a small portion of traffic to the new version
D
Create a Canary Deployment with Cloud Build to gradually shift traffic
E
Use a separate endpoint for the new model version and manually route traffic
No comments yet.