Query pattern–based data processing and storage optimization
Kyvos now supports query pattern-based data processing, in which aggregates are created based on actual user query patterns rather than pre-processing all columns defined in the semantic model. This approach significantly reduces semantic model processing time and storage costs by materializing only the aggregations that are actively used. New queries for which smart aggregates are not available are automatically served directly from the source data.
To enable query-based aggregate processing for a semantic model, perform the following steps.
Perform the following while scheduling a semantic model job. For more details on how to schedule a semantic model job, see the Scheduling semantic model processes section.
From the Materialize section, select the Metadata and Data option. Selecting this option will populate both the dimension metadata cache and data (Kyvos Analytical Store). Queries will be served from the Kyvos Analytical Store.
Select the Query-Based Aggregates checkbox. Aggregates are materialized from actual user query patterns and optimized for performance based on real usage and adaptable to evolving query behavior.
Uncheck the Raw Data checkbox.
