|
|
|
Enhancements have been made to the internal performance that concern aggregate sort nodes, composite histograms, and optimizer support for hash joins.
|
|
|
Improvements to aggregate handling allows Ingres to better support data-mining products such as CleverPath OLAP, which make extensive use of data aggregation.
Previous versions required a sort before doing grouping and aggregation. Ingres 2.6 now does grouping with hash bucketing instead of sorting. Hash grouping is usually faster than sorting. Other internal refinements streamline the calculation of common aggregates, reducing the amount of CPU time needed.
|
|
|
The composite histograms enhancement allows the creation of composite or multi-column histograms that model much more accurately the dependence of the values of one column on another, and lead to far better selectivity estimates and, ultimately, to better query plans.
|
|
|
Hash joins have been implemented in Ingres 2.6. A hash join is one in which a hash table is built with the rows of one of the join sources by hashing on the key columns of the join. The rows of the other join source are then read and hashed into the table on their key columns. The hashing of the second set of rows quickly identifies pairs of joining rows. This technique requires no index structures on the join columns (as does KEY join), nor does it require sorting on the join columns (as does merge join).