US 2026/0277902 A1 · Statistics For In-Memory Graph-Based Vector Index In A RDBMS Using A Cost-Based Optimizer

Abstract

A system for collecting statistics for an in-memory graph-based vector index executes a plurality of sample queries based on a set of search parameter values, a set of sampling filter ratios, and a set of sample vectors to perform a similarity search of a vector store using the in-memory graph-based vector index. The system determines, for each sample query, at least one cost statistic based on the execution and stores the cost statistics for a plurality of sample queries. A database system executes a vector query with a filter condition to search the vector store. The database system executes the vector query by determining a cost estimation for graph-based index access based at least in part on the set of cost statistics and executing the vector query based at least in part on the cost estimation for graph-based index access.

Claims20

Dependent claims

A method comprising:

collecting statistics for an in-memory graph-based vector index, wherein collecting the statistics comprises:
executing a plurality of sample queries based on a set of search parameter values, a set of sampling filter ratios, and a set of sample vectors to perform a similarity search of a vector store using the in-memory graph-based vector index;
determining, for each sample query within the plurality of sample queries, at least one cost statistic based on the execution;
and
storing the cost statistics for the plurality of sample queries;
executing, by a database system, a vector query with a filter condition to search the vector store, wherein executing the vector query comprises:
determining a cost estimation for graph-based index access based at least in part on the set of cost statistics;
and
executing the vector query based at least in part on the cost estimation for graph-based index access,
wherein the method is performed by one or more computing devices.

7 more elements

One or more non-transitory computer-readable media storing instructions which, when executed by one or more processors, cause:

collecting statistics for an in-memory graph-based vector index, wherein collecting the statistics comprises:
executing a plurality of sample queries based on a set of search parameter values, a set of sampling filter ratios, and a set of sample vectors to perform a similarity search of a vector store using the in-memory graph-based vector index;
determining, for each sample query within the plurality of sample queries, at least one cost statistic based on the execution;
and
storing the cost statistics for the plurality of sample queries;
executing, by a database system, a vector query with a filter condition to search the vector store, wherein executing the vector query comprises:
determining a cost estimation for graph-based index access based at least in part on the set of cost statistics;
and
executing the vector query based at least in part on the cost estimation for graph-based index access.

6 more elements

Dependencies are inferred from the claim text.

Statistics For In-Memory Graph-Based Vector Index In A RDBMS Using A Cost-Based Optimizer

Life of this patent

Statistics For In-Memory Graph-Based Vector Index In A RDBMS Using A Cost-Based Optimizer

Life of this patent

Abstract

Claims20