Comments on: Left-deep trees http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/ Just another Oracle weblog Tue, 18 Jun 2013 21:55:22 +0000 hourly 1 http://wordpress.com/ By: Kyle Hailey http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-36365 Sun, 30 May 2010 10:14:48 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-36365 For those asking, here is an specific example I ran into that shows a bushy plan to be much less expensive than a left deep
http://sites.google.com/site/embtdbo/sql-tuning-1#TOC-Query-3
(main point of post was showing Visual SQL Tuning but the result was rewriting the query into two subqueries with NO_MERGE hint

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-32736 Sun, 22 Mar 2009 00:59:04 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-32736 Smitha

It is perfectly feasible to find bushy tree plans that perform (much) better than left-deep trees. This is what makes the no_merge and push_pred hints so very important.

]]> By: Smitha http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-32718 Wed, 18 Mar 2009 11:42:47 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-32718 I have seen cases where in a bushy tree based query evaluation is much better than a left deep tree based one. Am I right? Please correct me if I am wrong

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-3546 Fri, 23 Feb 2007 07:58:36 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-3546 Krishna, I don’t think the optimizer can generate “bushy-tree” plans. It will always try to transform to a “left-deep”.

The only option you have is to use hints – specifically the /*+ no_merge */ hint, or subquery factoring with the /*+ materialize */ hint.

I will be writing about this in the future as part of my mini-series on Join Ordering.

]]> By: Krishna R http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-3543 Fri, 23 Feb 2007 07:43:44 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-3543 Hi, Can you please explain further how to go about improving performance of SQLs that need a bushy tree execution plan? I have been trying to tune some SQLs, and more or less all the badly performing ones needed bushy trees by nature.

Is hints the only option? Or can CBO create nice bushy execution plans if we have corresponding composite indexes?

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2484 Fri, 26 Jan 2007 03:33:44 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2484 Amihay, in general you will not get “bushy trees” unless you hint them. The optimizer always tries to transform your query to a “left-deep” tree.

Interesting idea about the parallelism – but I don’t think it can work like that. However, the example I ran in parallel did exhibit some interesting “out of order” behaviour.

]]> By: amihay gonen http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2384 Thu, 25 Jan 2007 11:23:55 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2384 Hi , in want cases we will have bushy tree ? does it mean that in those cases parallel execution code help us (one for each side ) ?

]]> By: Graham http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2378 Thu, 25 Jan 2007 10:09:32 +0000 http://jonathanlewis.wordpress.com/2007/01/24/left-deep-trees/#comment-2378 Thanks for the explanation Jonathan, all has become clear.

]]>