| 摘要 |
Systems, methods and program products for computing frequent value statistics such as frequent values in a column in a partitioned relational database, that may be employed across a number of nodes of a partitioned relational database and a case where a column is the single partitioning key of the database table as well as the case where the column is not the single partitioning key of the database table. The technique involves hashing column entries to a hash data set on a node, collecting the hash entries into another hash data set on a coordinating node into a combined hash data set, hashing entries from a table portion residing on the coordinating node into the combined hash data set and computing frequent value from the combined hash data set. Each hash data set is an array of counts and the combined hash data set is a combined array of counts.
|
