如何降低索引的clustering_factor

如何降低索引的clustering_factor 2012-06-08 22:40:19

分类： Oracle

 

环境:

OS:Red Hat Linux As 5

DB:10.2.0.4

 

我们知道判断一个索引的好坏可以通过该索引的clustering_factor高低来衡量,clustering_factor越低,索引的使用效果就越好,那怎么样才能降低索引的clustering_factor呢,通常使用的方法是让表的索引字段值按顺序存储,下面通过一个例子说明.

 

1.创建表并构造无序的数据

create table scott.tb_index_test
(
  id number not null,
  name varchar2(30)
);

 

create index scott.idx_tb_index_test  on scott.tb_index_test(id);

 

declare
  l_random_value number;
begin
  for i in  1 .. 100000 loop
    l_random_value := round(dbms_random.value(1,100000));
    insert into scott.tb_index_test(id) values (l_random_value);
    commit;
  end loop;
end;

 

2.分析表

 

begin
  dbms_stats.gather_table_stats(ownname => 'SCOTT',
                                tabname => 'TB_INDEX_TEST',
                                cascade => true);
end;

 

3.查看索引当前的clustering_factor

 

SQL> select ui.clustering_factor, ui.num_rows, ui.index_type, ui.distinct_keys
  from dba_indexes ui where ui.table_name = 'TB_INDEX_TEST';

CLUSTERING_FACTOR   NUM_ROWS INDEX_TYPE                  DISTINCT_KEYS
----------------- ---------- --------------------------- -------------
            99742     100000 NORMAL                             100000

 

4.将表中的数据按照索引字段存储

 

create table scott.tmp as select * from scott.tb_index_test;
truncate table scott.tb_index_test;
insert into scott.tb_index_test select * from scott.tmp order by id;

 

5.再次分析表

 

begin
  dbms_stats.gather_table_stats(ownname => 'SCOTT',
                                tabname => 'TB_INDEX_TEST',
                                cascade => true);
end;

 

6.这个时候索引的clustering_factor明显降低了

 

SQL> select ui.clustering_factor, ui.num_rows, ui.index_type, ui.distinct_keys
  from dba_indexes ui where ui.table_name = 'TB_INDEX_TEST';

CLUSTERING_FACTOR   NUM_ROWS INDEX_TYPE                  DISTINCT_KEYS
----------------- ---------- --------------------------- -------------
              372     100000 NORMAL                             100000

 

-- The End --

转载于:https://www.cnblogs.com/weixun/archive/2013/03/27/2985567.html