You need to label entities in viewed videos and provide very fast filtering on several TB of user-preference data. Which design meets both needs?

The correct answer leverages Cloud Video Intelligence API for efficient, pre-trained video labeling, which is a common and effective solution for this task. Storing the labeled data in Cloud Bigtable is ideal for the "very fast filtering on several TB of user-preference data" requirement. Bigtable is a wide-column NoSQL database designed for high-throughput, low-latency access to large datasets, making it perfect for real-time filtering and recommendations. Incorrect options: Training a complex Spark MLlib classifier and deploying on Dataproc is an option for custom labeling, but it's more resource-intensive and time-consuming than using a pre-trained API like Video Intelligence for a common task. It also doesn't address the "very fast filtering" requirement as effectively as Bigtable. Training two Spark MLlib models has similar drawbacks regarding complexity and efficiency compared to using a specialized API and Bigtable. Using Cloud SQL for storing results is less suitable for "very fast filtering on several TB" than Bigtable. Cloud SQL is a relational database optimized for structured data and complex queries, but it typically doesn't offer the same low-latency, high-throughput performance as Bigtable for massive-scale key-value or wide-column lookups.