Baoding-FenShuaJiang
/
Green_GoodERP18_FSJ_Standard
Segui 1
Vota 0
Wiki 捐赠
gooderp18绿色标准版
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Albero (Tree): 0f75e31513
Rami (Branch) Tag
${ item.name }
Crea branch ${ searchTerm }
da '0f75e31513'
${ noResults }
Green_GoodERP18_FSJ_Standard/PostgreSQL/doc/postgresql/html/geqo-pg-intro.html

107 lines
9.2KB
Originale Blame Cronologia 

  1. <?xml version="1.0" encoding="UTF-8" standalone="no"?>

  2. <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>59.3. Genetic Query Optimization (GEQO) in PostgreSQL</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@lists.postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets V1.79.1" /><link rel="prev" href="geqo-intro2.html" title="59.2. Genetic Algorithms" /><link rel="next" href="geqo-biblio.html" title="59.4. Further Reading" /></head><body><div xmlns="http://www.w3.org/TR/xhtml1/transitional" class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">59.3. Genetic Query Optimization (<acronym xmlns="http://www.w3.org/1999/xhtml" class="acronym">GEQO</acronym>) in PostgreSQL</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="geqo-intro2.html" title="59.2. Genetic Algorithms">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="geqo.html" title="Chapter 59. Genetic Query Optimizer">Up</a></td><th width="60%" align="center">Chapter 59. Genetic Query Optimizer</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 12.4 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="geqo-biblio.html" title="59.4. Further Reading">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="GEQO-PG-INTRO"><div class="titlepage"><div><div><h2 class="title" style="clear: both">59.3. Genetic Query Optimization (<acronym class="acronym">GEQO</acronym>) in PostgreSQL</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="geqo-pg-intro.html#id-1.10.12.5.6">59.3.1. Generating Possible Plans with <acronym class="acronym">GEQO</acronym></a></span></dt><dt><span class="sect2"><a href="geqo-pg-intro.html#GEQO-FUTURE">59.3.2. Future Implementation Tasks for

  3.     <span class="productname">PostgreSQL</span> <acronym class="acronym">GEQO</acronym></a></span></dt></dl></div><p>

  4.     The <acronym class="acronym">GEQO</acronym> module approaches the query

  5.     optimization problem as though it were the well-known traveling salesman

  6.     problem (<acronym class="acronym">TSP</acronym>).

  7.     Possible query plans are encoded as integer strings. Each string

  8.     represents the join order from one relation of the query to the next.

  9.     For example, the join tree

  10. </p><pre class="literallayout">

  11.    /\

  12.   /\ 2

  13.  /\ 3

  14. 4  1

  15. </pre><p>

  16.     is encoded by the integer string '4-1-3-2',

  17.     which means, first join relation '4' and '1', then '3', and

  18.     then '2', where 1, 2, 3, 4 are relation IDs within the

  19.     <span class="productname">PostgreSQL</span> optimizer.

  20.    </p><p>

  21.     Specific characteristics of the <acronym class="acronym">GEQO</acronym>

  22.     implementation in <span class="productname">PostgreSQL</span>

  23.     are:

  24. 

  25.     </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: bullet; "><li class="listitem" style="list-style-type: disc"><p>

  26.        Usage of a <em class="firstterm">steady state</em> <acronym class="acronym">GA</acronym> (replacement of the least fit

  27.        individuals in a population, not whole-generational replacement)

  28.        allows fast convergence towards improved query plans. This is

  29.        essential for query handling with reasonable time;

  30.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  31.        Usage of <em class="firstterm">edge recombination crossover</em>

  32.        which is especially suited to keep edge losses low for the

  33.        solution of the <acronym class="acronym">TSP</acronym> by means of a

  34.        <acronym class="acronym">GA</acronym>;

  35.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  36.        Mutation as genetic operator is deprecated so that no repair

  37.        mechanisms are needed to generate legal <acronym class="acronym">TSP</acronym> tours.

  38.       </p></li></ul></div><p>

  39.    </p><p>

  40.     Parts of the <acronym class="acronym">GEQO</acronym> module are adapted from D. Whitley's

  41.     Genitor algorithm.

  42.    </p><p>

  43.     The <acronym class="acronym">GEQO</acronym> module allows

  44.     the <span class="productname">PostgreSQL</span> query optimizer to

  45.     support large join queries effectively through

  46.     non-exhaustive search.

  47.    </p><div class="sect2" id="id-1.10.12.5.6"><div class="titlepage"><div><div><h3 class="title">59.3.1. Generating Possible Plans with <acronym class="acronym">GEQO</acronym></h3></div></div></div><p>

  48.     The <acronym class="acronym">GEQO</acronym> planning process uses the standard planner

  49.     code to generate plans for scans of individual relations.  Then join

  50.     plans are developed using the genetic approach.  As shown above, each

  51.     candidate join plan is represented by a sequence in which to join

  52.     the base relations.  In the initial stage, the <acronym class="acronym">GEQO</acronym>

  53.     code simply generates some possible join sequences at random.  For each

  54.     join sequence considered, the standard planner code is invoked to

  55.     estimate the cost of performing the query using that join sequence.

  56.     (For each step of the join sequence, all three possible join strategies

  57.     are considered; and all the initially-determined relation scan plans

  58.     are available.  The estimated cost is the cheapest of these

  59.     possibilities.)  Join sequences with lower estimated cost are considered

  60.     <span class="quote">“<span class="quote">more fit</span>”</span> than those with higher cost.  The genetic algorithm

  61.     discards the least fit candidates.  Then new candidates are generated

  62.     by combining genes of more-fit candidates — that is, by using

  63.     randomly-chosen portions of known low-cost join sequences to create

  64.     new sequences for consideration.  This process is repeated until a

  65.     preset number of join sequences have been considered; then the best

  66.     one found at any time during the search is used to generate the finished

  67.     plan.

  68.    </p><p>

  69.     This process is inherently nondeterministic, because of the randomized

  70.     choices made during both the initial population selection and subsequent

  71.     <span class="quote">“<span class="quote">mutation</span>”</span> of the best candidates.  To avoid surprising changes

  72.     of the selected plan, each run of the GEQO algorithm restarts its

  73.     random number generator with the current <a class="xref" href="runtime-config-query.html#GUC-GEQO-SEED">geqo_seed</a>

  74.     parameter setting.  As long as <code class="varname">geqo_seed</code> and the other

  75.     GEQO parameters are kept fixed, the same plan will be generated for a

  76.     given query (and other planner inputs such as statistics).  To experiment

  77.     with different search paths, try changing <code class="varname">geqo_seed</code>.

  78.    </p></div><div class="sect2" id="GEQO-FUTURE"><div class="titlepage"><div><div><h3 class="title">59.3.2. Future Implementation Tasks for

  79.     <span class="productname">PostgreSQL</span> <acronym class="acronym">GEQO</acronym></h3></div></div></div><p>

  80.       Work is still needed to improve the genetic algorithm parameter

  81.       settings.

  82.       In file <code class="filename">src/backend/optimizer/geqo/geqo_main.c</code>,

  83.       routines

  84.       <code class="function">gimme_pool_size</code> and <code class="function">gimme_number_generations</code>,

  85.       we have to find a compromise for the parameter settings

  86.       to satisfy two competing demands:

  87.       </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: disc; "><li class="listitem"><p>

  88.          Optimality of the query plan

  89.         </p></li><li class="listitem"><p>

  90.          Computing time

  91.         </p></li></ul></div><p>

  92.      </p><p>

  93.       In the current implementation, the fitness of each candidate join

  94.       sequence is estimated by running the standard planner's join selection

  95.       and cost estimation code from scratch.  To the extent that different

  96.       candidates use similar sub-sequences of joins, a great deal of work

  97.       will be repeated.  This could be made significantly faster by retaining

  98.       cost estimates for sub-joins.  The problem is to avoid expending

  99.       unreasonable amounts of memory on retaining that state.

  100.      </p><p>

  101.       At a more basic level, it is not clear that solving query optimization

  102.       with a GA algorithm designed for TSP is appropriate.  In the TSP case,

  103.       the cost associated with any substring (partial tour) is independent

  104.       of the rest of the tour, but this is certainly not true for query

  105.       optimization.  Thus it is questionable whether edge recombination

  106.       crossover is the most effective mutation procedure.

  107.      </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="geqo-intro2.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="geqo.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="geqo-biblio.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">59.2. Genetic Algorithms </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 59.4. Further Reading</td></tr></table></div></body></html>
上海开阖软件有限公司 沪ICP备12045867号-1