Labatt
50-Minute Talk
Optimizer statistics currently consist of two types: regular statistics concerning a given column in a table, and extended statistics which cover relationships between combinations of a defined set of columns in a given table. Currently there is no way to cover relationships between columns in two different tables, and no way to filter and weigh statistics in one table based upon how frequently the rows are joined to another table.
This talk will cover the potential value of such statistics for query planning, a proposed data model for storing such statistics, and how those statistics would actually be collected and maintained.
