I. Features Introduced in Ingres II 2.5 : Optimizer and Optimizedb Enhancements
 
Share this page                  
Optimizer and Optimizedb Enhancements
A variety of enhancements have been made to optimizedb.
The flag -zlr causes optimizedb to retain the original repetition factor when rebuilding an existing histogram. This is useful when a histogram (and its repetition factor) is built once by reading the whole table, then updated later using sampling (which can produce inaccurate repetition factors).
A minor bug was fixed to allow the ā€œlā€ flag to request an exclusive lock on the database during optimizedb processing (just as for other command line utilities).
The current limit of 1000 parameters coded in a separate file using the -zf parameter has been lifted. There is now no limit to the number of such parameters.
An -o filename option (similar to that in statdump) has been added to optimizedb. It creates a statdump-style output file, which can then be input back to optimizedb with the -i filename option. But more importantly, it does not update the catalog iistatistics and iihistogram tables. This allows a busy shop to construct histograms at anytime, with no worry about update conflicts. Then at a convenient later time, optimizedb can be run with the -i option to add the histograms to the catalog.
In addition to the flag enhancements, an enhancement has been introduced to allow more accurate histograms to be built on columns with significant skew in their value sets. Specifically, a column with many distinct values, which would generate an inexact histogram, now produces exact cells for values that occur significantly more than the average. This permits much more accurate estimates in the compilation of queries with restrictions on such columns, and, therefore, better query plans.
The query compiler has been enhanced to compile more efficient strategies for complex queries involving aggregate views and unions.