db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mamta A. Satoor (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (DERBY-1907) Improve optimizer to prune query plans more aggressively.
Date Tue, 25 Jun 2013 05:06:22 GMT

     [ https://issues.apache.org/jira/browse/DERBY-1907?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Mamta A. Satoor updated DERBY-1907:

    Urgency: Normal
     Labels: derby_triage10_11  (was: )
> Improve optimizer to prune query plans more aggressively.
> ---------------------------------------------------------
>                 Key: DERBY-1907
>                 URL: https://issues.apache.org/jira/browse/DERBY-1907
>             Project: Derby
>          Issue Type: Improvement
>          Components: SQL
>    Affects Versions:,,
>            Reporter: A B
>              Labels: derby_triage10_11
> For large queries with numerous and deeply-nested subqueries like those in DERBY-1777,
the number of different index/joinOrder/joinStrategy combinations that the optimizer can process
is extremely high.  In the absence of an efficient timeout mechanism (see DERBY-1906) this
can lead to situations where the optimizer spends far too much time trying out different access
paths before picking a plan.  In the case of the DERBY-1777 queries, the optimizer can literally
spend hours trying out all of the possible combinations before deciding on the best one.
> Even if/when the timeout mechanism is improved, it would still be great if the Derby
optimizer could somehow figure out how to aggressively prune out access paths that are known
(or suspected) to be sub-optimal.  This way the optimizer would spend less time optimizing
"bad" access paths and could therefore try out more "good" plans before timeout occurs.
> I have no idea how this pruning would occur, but I do think it would be extremely beneficial
to the Derby optimization process when dealing with deeply nested queries with large FROM

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message