1Z0-184-25 Question 61
Single answerYou are designing a high-traffic image-sharing application in Oracle Cloud Infrastructure that must handle large-scale similarity searches on user-uploaded images. You decide to use an IVF (Inverted File) vector index to facilitate approximate nearest-neighbor queries on image embeddings. The dataset is sizeable, and queries need to return results quickly without excessively long index build times. Which approach helps optimize query performance while controlling the overhead of maintaining the IVF vector index?
- A
Configure a very large number of IVF centroids to maximize recall, regardless of index build time.
- B
Choose a balanced number of IVF centroids that suits the dataset size and tune the query parameters to reduce search complexity.
- C
Disable periodic index maintenance and rebuild the index fully once a week to ensure consistent performance.
- D
Use minimal centroids and rely on full vector scans to ensure you do not miss any potential match.
Show answer and explanation
Correct answer: B
Explanation
IVF (Inverted File) vector indexing in OCI relies on dividing high-dimensional vector data into clusters (centroids), which reduces search time by focusing on relevant clusters. Oracle best practices recommend balancing the number of centroids with the dataset size and query frequency. Too many centroids can inflate build times and indexing overhead, while too few centroids can degrade performance by forcing near-full scans. Periodic and incremental index maintenance also ensures fresh, accurate query results in fast-paced data environments. Refer to the Oracle Cloud Infrastructure documentation on managing vector indexes for specific configuration guidelines.
- A. Incorrect.
Option 1 is incorrect. Creating a very large number of centroids (clusters) can offer higher recall but significantly increases index build time and can degrade query performance if the centroids are not well-tuned. It's not always best to maximize recall without balancing cost and query latency.
- B. Correct.
Option 2 is correct. A balanced approach�choosing an appropriate number of IVF centroids based on dataset size and leveraging tunable search parameters such as probe counts�helps control index build time while preserving good query performance. This best practice is recommended in Oracle documentation for scalable approximate nearest-neighbor tasks.
- C. Incorrect.
Option 3 is incorrect because postponing all index maintenance to infrequent intervals can cause the index to become stale. This leads to lower relevance for newly ingested data and potentially inconsistent performance across the week.
- D. Incorrect.
Option 4 is incorrect. Using very few centroids essentially undermines the advantage of the IVF approach, turning queries into nearly full vector scans, which can be slow and negates the benefit of clustering for approximate nearest-neighbor searches.