1Z0-184-25 Question 41
Single answerYour team has created product embeddings in an Autonomous Database on Oracle Cloud Infrastructure to power a recommendation engine. You have built a vector index on the embeddings column to support similarity queries, but you�ve noticed increased resource usage when multiple concurrent queries run. Which approach will most effectively optimize concurrency and overall performance of vector searches while managing costs?
- A
Partition the vector index based on product categories to reduce the search space across concurrent queries
- B
Increase the dimension of the vectors to capture more features for each product embedding
- C
Rely on a standard B-tree index instead of a vector index to handle similar product searches
- D
Store the embeddings in an Object Storage bucket and query them directly without indexing
Show answer and explanation
Correct answer: A
Explanation
To handle large-scale concurrent queries on product embeddings, partitioning the vector index is a common best practice in Oracle Autonomous Database. It confines the search scope, enabling faster lookups while distributing the workload across partitions. For details on managing high-concurrency vector searches, refer to the official Oracle documentation on vector indexes and indexing strategies in Autonomous Database.
- A. Correct.
Partitioning the vector index based on product categories or another logical segmentation helps reduce the search space for each query. With smaller partitions, the index can handle concurrent searches more efficiently and reduce query response time, helping keep resource utilization in check.
- B. Incorrect.
Increasing the vector dimension typically increases both storage requirements and query complexity, which can degrade performance rather than improve it. This approach may be chosen for improved accuracy in some scenarios, but it won't generally reduce resource usage under high concurrency.
- C. Incorrect.
A standard B-tree index is not optimized for similarity searches on high-dimensional embeddings. Such indexes are appropriate for exact matches or range queries on numeric or string data, not for retrieving nearest neighbors in a vector space.
- D. Incorrect.
Object Storage does not provide a native vector indexing or querying mechanism. While storing data there is cost-effective for static files, it does not help perform similarity queries in real time or reduce concurrency-related overhead.