SQ4
A team needs to serve a custom TensorFlow model with autoscaling, A/B traffic splitting, and monitoring. Which is the best option?
- ADeploy on a single Compute Engine VM
- BUse Cloud Run with the model bundled in a container
- CServe from Cloud Storage signed URLs
- DUse Vertex AI Endpoints
Show answerHide answer
Answer: D
Vertex AI Endpoints provide managed model serving with autoscaling, traffic splitting for A/B testing, and integrated monitoring. Cloud Run could work for simpler cases but lacks native ML serving features like traffic splitting between model versions.