Explanation:
The optimal configuration for a real-time anomaly detection system with the given requirements involves using Cloud Pub/Sub for data ingestion to handle high-velocity data streams, Dataflow for efficient stream processing with low latency, BigQuery for scalable and cost-effective storage enabling analytics and visualization, and Vertex AI for developing and deploying the machine learning model. This setup ensures scalability, cost-effectiveness, and compliance with data security standards. Options A and C both correctly identify the use of Dataflow and Vertex AI, but option A more accurately specifies the use of BigQuery for storage, which is better suited for analytics and visualization purposes compared to Cloud Storage or Cloud Bigtable mentioned in other options.
Ultimate access to all questions.
No comments yet.
You are a Machine Learning Engineer at a manufacturing company tasked with developing a real-time anomaly detection system for sensor data across multiple production lines. The system must process high-velocity data streams, identify anomalies with low latency, and store the results for further analytics and visualization. The solution must be cost-effective, scalable, and comply with industry standards for data security. Given these requirements, how would you best architect the pipeline using Google Cloud services? Choose the two most appropriate options. (Choose two)
A
Use Cloud Pub/Sub for ingesting real-time sensor data, Dataflow for stream processing, BigQuery for storing processed data, and Vertex AI for training and deploying the anomaly detection model.
B
Use Cloud Pub/Sub for data ingestion, Cloud Functions for processing each message individually, Cloud Storage for storing results, and AutoML for model training and deployment.
C
Use Cloud Pub/Sub for data ingestion, Dataflow for real-time processing, BigQuery for analytics storage, and Vertex AI for the machine learning model.
D
Use Cloud Pub/Sub for data ingestion, Dataproc for processing, Cloud Bigtable for storage, and AutoML for the machine learning model.
E
All of the above options are suitable for different scenarios depending on the specific requirements and constraints.