Horizontal interval-based data partitioning and indexing for large clusters
US8903803B1 · kind B1 · utility
Assignee
Inventors
Key dates
| Filing date | Jun 18, 2014 |
| Grant date | Dec 2, 2014 |
| Priority date | — |
| Expiry date | Jun 18, 2034 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06F16/2264
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Current data records having a start and end time are transformed into a 2D space having a first dimension for each data record's start time and a second dimension for each data record's end time. Historical queries specifying data ranges are obtained. A response was previously sent for each historical query and specifying a sub-portion of data records that overlap with such historical query's specified data range. Partitioning schemes for the current data records in the 2D space are generated. An optimum partitioning scheme having a lowest cost is selected based on costs of executing the historical queries with respect to each of the partitioning schemes. The optimum partitioning scheme is applied on the current data records, including newly received data records, in the 2D space so that any subsequently received queries are applied against the current data records as partitioned by the optimum partitioning scheme in the 2D space.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.