3.1.2 AI Smart Scan Enhancements

Oracle AI Database includes artificial intelligence (AI) vector search capabilities designed to efficiently query data based on semantic similarities. Oracle AI Vector Search enables applications like voice assistants, chatbots, language translators, recommendation systems, and anomaly detection systems. AI Vector Search also facilitates sophisticated search and recognition capabilities on images and videos.

Complementing Oracle AI Vector Search, Oracle Exadata System Software includes AI Smart Scan, a collection of Exadata-specific optimizations capable of improving the performance of various AI vector query operations by orders of magnitude.

AI Smart Scan automatically accelerates Oracle AI Vector Search with optimizations that deliver low-latency parallelized scans across massive volumes of vector data. It processes vector data at memory speed, leveraging ultra-fast Exadata RDMA Memory (XRMEM) and Exadata Smart Flash Cache in the Exadata storage servers. AI Smart Scan also performs vector distance computations and top-K filtering at the data source, avoiding unnecessary network data transfer and database server processing.

Oracle Exadata System Software release 25.2.0 includes the following AI Smart Scan enhancements:

• Optimized support for neighbor partition vector indexes with included columns.

  Previously, queries combining business data and AI vector searches required reads on the base table to evaluate relational predicates. For example, in the following query, rows returned by the vector search still require additional reads on the houses table to apply the filters on price and zipcode, and to retrieve the address and price values.

  SQL> SELECT address, price
       FROM houses
       WHERE price < 950000 and zipcode = 94065
       ORDER BY VECTOR_DISTANCE(description_vector, :query_vector)
       FETCH APPROX FIRST 10 ROWS ONLY;

  Starting with Oracle AI Database release 23.7, neighbor partition vector indexes can include additional columns containing business data. For example:

  SQL> CREATE VECTOR INDEX house_idx ON houses(description_vector)
       INCLUDE (address, price, zipcode)
       ORGANIZATION NEIGHBOR PARTITIONS;

  With Oracle Exadata System Software release 25.2.0, AI Smart Scan can process these types of queries using only the vector index, with the included relational columns, which significantly improves overall query performance by eliminating the base table lookups.

• Sparse vectors

  In conjunction with Oracle AI Database release 23.7 or later, Oracle Exadata System Software release 25.2.0 enables AI Smart Scan using sparse vectors.

  Sparse vectors typically have many dimensions but contain very few non-zero values. Sparse encoding models, such as SPLADE and BM25, generate sparse vectors. In suitable applications, such as natural language processing, sparse vectors offer considerable performance advantages, enabling faster query processing while reducing memory and CPU usage.