1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
5 >Genetic Query Optimization (GEQO) in PostgreSQL</TITLE
8 CONTENT="Modular DocBook HTML Stylesheet Version 1.7"><LINK
10 HREF="mailto:pgsql-docs@postgresql.org"><LINK
12 TITLE="PostgreSQL 7.4.1 Documentation"
13 HREF="index.html"><LINK
15 TITLE="Genetic Query Optimizer"
16 HREF="geqo.html"><LINK
18 TITLE="Genetic Algorithms"
19 HREF="geqo-intro2.html"><LINK
21 TITLE="Further Readings"
22 HREF="geqo-biblio.html"><LINK
25 HREF="stylesheet.css"><META
27 CONTENT="2003-12-22T03:48:47"></HEAD
33 SUMMARY="Header navigation table"
43 >PostgreSQL 7.4.1 Documentation</TH
51 HREF="geqo-intro2.html"
67 >Chapter 48. Genetic Query Optimizer</TD
81 HREF="geqo-biblio.html"
96 >48.3. Genetic Query Optimization (<ACRONYM
105 > module is intended for the solution of the query
106 optimization problem similar to a traveling salesman problem (<ACRONYM
110 Possible query plans are encoded as integer strings. Each string
111 represents the join order from one relation of the query to the next.
112 E. g., the query tree
114 CLASS="LITERALLAYOUT"
120 is encoded by the integer string '4-1-3-2',
121 which means, first join relation '4' and '1', then '3', and
122 then '2', where 1, 2, 3, 4 are relation IDs within the
129 > Parts of the <ACRONYM
132 > module are adapted from D. Whitley's Genitor
136 > Specific characteristics of the <ACRONYM
140 implementation in <SPAN
151 STYLE="list-style-type: disc"
159 > (replacement of the least fit
160 individuals in a population, not whole-generational replacement)
161 allows fast convergence towards improved query plans. This is
162 essential for query handling with reasonable time;
166 STYLE="list-style-type: disc"
170 >edge recombination crossover</I
172 which is especially suited to keep edge losses low for the
173 solution of the <ACRONYM
184 STYLE="list-style-type: disc"
186 > Mutation as genetic operator is deprecated so that no repair
187 mechanisms are needed to generate legal <ACRONYM
205 support large join queries effectively through
206 non-exhaustive search.
214 >48.3.1. Future Implementation Tasks for
224 > Work is still needed to improve the genetic algorithm parameter
228 >backend/optimizer/geqo/geqo_params.c</TT
232 >gimme_pool_size</CODE
235 >gimme_number_generations</CODE
237 we have to find a compromise for the parameter settings
238 to satisfy two competing demands:
245 > Optimality of the query plan
263 SUMMARY="Footer navigation table"
274 HREF="geqo-intro2.html"
292 HREF="geqo-biblio.html"
302 >Genetic Algorithms</TD
316 >Further Readings</TD