Answer-first summary for fast verification
Answer: Deploy a Dataflow streaming pipeline with the Runlnference API, and use automatic model refresh.
The question requires a solution that supports 24/7 operation without downtime, sub-millisecond predictions, daily model retraining, and cost-effectiveness. Option D (Dataflow streaming pipeline with RunInference API and automatic model refresh) is optimal because: 1) Dataflow streaming enables continuous, real-time processing of sensor data, meeting the 24/7 and low-latency requirements. 2) The RunInference API allows direct model invocation within the pipeline, eliminating the need for external endpoints like Vertex AI Prediction, which reduces latency and costs. 3) Automatic model refresh seamlessly updates models daily without pipeline downtime, aligning with the retraining schedule. In contrast, options A and B use batch pipelines, which are unsuitable for real-time predictions. Option C uses a Vertex AI Prediction endpoint, which incurs higher costs and potential latency compared to RunInference, and autoscaling may introduce brief downtime during replica updates, violating the no-downtime requirement. Community discussion strongly supports D (74% consensus), highlighting its cost-effectiveness, low latency, and seamless model updates.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
No comments yet.
You work for a textile manufacturing company with hundreds of machines, each equipped with multiple sensors. Your team has built hundreds of machine learning models using this sensor data to detect machine anomalies. These models are retrained daily and need to be deployed in a cost-effective manner. The deployment must support 24/7 operation with no downtime and provide sub-millisecond prediction latency. What should you do?
A
Deploy a Dataflow batch pipeline and a Vertex AI Prediction endpoint.
B
Deploy a Dataflow batch pipeline with the Runlnference API, and use model refresh.
C
Deploy a Dataflow streaming pipeline and a Vertex AI Prediction endpoint with autoscaling.
D
Deploy a Dataflow streaming pipeline with the Runlnference API, and use automatic model refresh.