Cardinality Estimation Applying Micro Self-Tuning Histogram

In the cardinality estimation solutions based on multi-dimensional self-tuning histograms, periodical data scans are avoided and self-tuning histograms are constructed according to query feedback records. We call this kind of cardinality estimation solutions the reactive solutions. The existing reactive solutions are stuck with the issue of "curse of dimension". And they are unpredictable and time-consuming. To address these issues, a new reactive solution is proposed in the paper. A micro self-tuning histogram only covering the neighborhood of the new predicate is constructed, which is a beneficial attempt to improve the cardinality estimation efficiency under high dimensions, and notably alleviate the issue of "curse of dimension". Furthermore, the process of meeting a space budget is eliminated completely, which makes the whole solution reliable and dexterous.