INCREMENTAL CARDINALITY ESTIMATION FOR A SET OF DATA VALUES

A technique is provided for incrementally maintaining column cardinality estimates in database management systems. The system catalog table containing a cardinality estimate for a column is extended to include an appropriate data structure. A modified linear counting technique is used in a first embodiment of a method for column cardinality estimation. Moreover, a modified logarithmic counting technique is used in a second, preferred embodiment of a column cardinality estimation method to reduce storage requirements for the data structure. The cardinality estimate is produced by an initial scan of the data but is then further maintained without requiring a full scan of the data. Data changes are reflected incrementally in modifications to the initial cardinality estimate, keeping the cardinality statistics more current with respect to the database condition. The technique of the invention typically provides a capability for a database management system to produce more efficient search plans providing more effective responses to user queries through the use of improved cardinality statistics.