Techniques for ordering predicates in column partitioned databases for query optimization

A processor-implemented system, comprising:

• a database implemented and programmed within a non-transitory computer-readable storage medium, said database including database tables partitioned on columns, wherein when a base column partitioned (CP) table is accessed by the processor for evaluating a query, the processor accesses only column partitions that are referenced in the query;
• a memory programmed with a query optimizer for execution on a processor;
• the query optimizer is configured to reorganize predicates that access column partitions within a query based on selectivity and costs, and to process the reorganized query using a summary structure for all predicates on a single CP table that records a total cost to evaluate those predicates and a combined selectivity for those predicates, and each unique predicate grouped based on the column partitions into a predicate-column partition (CP) node having a particular summary structure, each column partition referenced within a particular predicate-CP node is linked to a column partition node which records a column partition index, a column type, a column width, and a compression ratio for that column partition and a total number of logical rows that has to be accessed with that column partition, the query optimizer further configured to structure references to the query, the predicates, the ordered groups, the column partitions, and selectivity and cost estimates within metadata carried with the query.