What are the decisions and rules that the statistical optimimizer uses ??
If you can't give out the details, how 'bout some general rules.
Any help would be greatly appreciated!
Chuck (cdm_at_idm.com)
Below is an explanation of why we are using statitsical optimization:
We have found that we need to use the statistical optimizer to get proper
nested loop plans with many of our queries using distibuted joins.
With the rule based optimizer, we would always get merge/sort joins in
distributed queries. When the remote table is over 70 Megs, a merge/sort
join will try to copy a huge chunk of the remote table into the local table
space. It just takes forever... We call situations like these 'Black Holes'.
We have found many, many queries in our product taht work fine in a single
database that turn into 'Black Holes' when we distribute the tables across
2 or more systems.