Warehouse Data Partition Index Management

The patent segments the data warehouse into multiple partitions and further divides affected partitions into subsets. Indexes are selectively disabled only on the specific subset of partitions being updated, rather than on all partitions or the entire data warehouse. This segmentation allows query performance to be maintained on unaffected partitions while reducing processing time on affected partitions during incremental loads.

The patent applies local quality by making indexes disabled only in the specific locations (partition subsets) where data is being loaded. The index state varies locally across different partitions - enabled on stable partitions for optimal query performance, and disabled only on the local subset being updated to minimize processing overhead. This localized approach resolves the contradiction between maintaining overall query performance and reducing local processing time.

The patent segments partitions into affected and unaffected groups, and further divides affected partitions into subsets. Indexes are disabled only on the specific subset being updated, while remaining enabled on all other partitions. This allows the system to achieve high loading speed on the affected subset without compromising query performance on the larger portion of the data warehouse where indexes remain enabled.

Instead of disabling indexes on all partitions (excessive action), the patent applies partial action by disabling indexes only on the specific subset of partitions currently being updated. This partial approach achieves sufficient loading speed improvement for the active workload while avoiding the performance penalty on the majority of partitions that do not require updating.

The patent extracts and utilizes metadata from data dictionaries that pre-describe the structure and partitioning of the data warehouse. By querying the data dictionary with the load specification, the system directly identifies affected partitions without scanning the entire data warehouse. This extraction approach maintains accurate partition identification while dramatically reducing processing overhead compared to full scans.

The data dictionary contains pre-computed metadata about partition structures, data locations, and relationships. This preliminary organization of information allows the system to quickly determine affected partitions through metadata queries rather than performing expensive runtime scans. The preliminary action of organizing data into partitioned structures with associated metadata enables efficient identification with minimal processing overhead.

The patent segments affected partitions into manageable subsets and disables bitmap indexes only on these specific subsets during loading. After loading completes, index rebuilding is performed only on the same limited subsets rather than on the entire data warehouse. This segmentation reduces both the loading performance penalty and the subsequent index maintenance time, as the workload is confined to small partition subsets rather than the full data warehouse.

The patent applies partial action by disabling and subsequently rebuilding indexes only on the subset of partitions that were updated, rather than performing full data warehouse index maintenance. This partial approach achieves sufficient loading performance improvement while minimizing the time lost to index rebuilding, as the maintenance workload is limited to the small subset of affected partitions rather than the entire data warehouse.