关于cursor_sharing = similar

类别:数据库 点击:0 评论:0 推荐:

本文作者: biti_rainy ([email protected])

 

摘要:本文通过简单实验来尝试说明cursor_sharing=similar的含义。

1.1.        实验现象

我们先看看在表没有分析无统计数据情况下的表现 

SQL> alter session set cursor_sharing = similar;

 

Session altered.

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4948

parse time elapsed                                                    4468

parse count (total)                                                   170148

parse count (hard)                                                    1619  (硬分析次数)

parse count (failures)                                                80

 

SQL> select count(*) from t where object_id = 1000;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4948

parse time elapsed                                                     4468

parse count (total)                                                  170172

parse count (hard)                                                     1620

parse count (failures)                                                   80

 

SQL> /

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4948

parse time elapsed                                                     4468

parse count (total)                                                  170176

parse count (hard)                                                     1620

parse count (failures)                                                   80

 

SQL> select count(*) from t where object_id = 1000;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4948

parse time elapsed                                                     4468

parse count (total)                                                  170178

parse count (hard)                                                     1620

parse count (failures)                                                   80

 

SQL> select count(*) from t where object_id = 1001;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4948

parse time elapsed                                                     4468

parse count (total)                                                  170180

parse count (hard)                                              1620(即使object_id发生变化依然没有硬解析)

parse count (failures)                                                   80

 

我们再来看分析表和字段信息后的表现

SQL> analyze table t1 compute statistics for table for columns object_id;

 

Table analyzed.

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4973

parse time elapsed                                                     4495

parse count (total)                                                  170982

parse count (hard)                                                     1640

parse count (failures)                                                   80

 

SQL> select count(*) from t1 where object_id = 5000;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4973

parse time elapsed                                                     4495

parse count (total)                                                  170984

parse count (hard)                                                     1641

parse count (failures)                                                   80

  

SQL> select count(*) from t1 where object_id = 5000;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

 

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4973

parse time elapsed                                                     4495

parse count (total)                                                  171008

parse count (hard)                                                     1641 (重复执行没发生变化)

parse count (failures)                                                   80

 

SQL> select count(*) from t1 where object_id = 5001;

 

   COUNT(*)

----------

          0

 

SQL> select name,value from v$sysstat where name like '%parse%';

  

NAME                                                                  VALUE

---------------------------------------------------------------- ----------

parse time cpu                                                         4973

parse time elapsed                                                     4495

parse count (total)                                                  171010

parse count (hard)                                                 1642 (当object_id变化的时候产生硬分析)

parse count (failures)                                                   80

 

SQL>

 

SQL> select sql_text,child_number from v$sql where sql_text like 'select count(*) from t1 where%';

 

SQL_TEXT

------------------------------------------------------------------------------

CHILD_NUMBER

------------

select count(*) from t1 where object_id = :"SYS_B_0"

            0

 

select count(*) from t1 where object_id = :"SYS_B_0"

            1

 

1.2.        结论

可以看出若存在object_id的 histograms ,则每次是不同的值的时候都产生硬解析 ,若不存在 histograms,则不产生硬解析。换句话说,当表的字段被分析过存在histograms的时候,similar 的表现和exact一样,当表的字段没被分析,不存在histograms的时候,similar的表现和force一样。这样避免了一味地如force一样转换成变量形式,因为有histograms的情况下转换成变量之后就容易产生错误的执行计划,没有利用上统计信息。而exact呢,在没有histograms的情况下也要分别产生硬解析,这样的话,由于执行计划不会受到数据分布的影响(因为没有统计信息)重新解析是没有实质意义的。而similar则综合了两者的优点。

 

 

作者简介:

网名coolyl

Oracle专家。有丰富的Oracle实践经验,对体系结构、备份恢复、SQL优化、整体性能优化、Oracle Internal都有深入研究。

个人教育和成长经历:

对数据库应用设计中如何正确地应用oracle特性以扬长避短具有深刻理解。

有丰富的oracle实践经验,对数据库的体系结构、备份恢复、sql优化、数据库整体性能优化、oracle internal都有深入研究。

曾于某电信集成公司负责计费系统的开发,然后成为某系统集成公司的DBA,再辗转在香港一家跨国公司珠海研发中心担任技术负责人(公司主要产品就是sql与数据库优化工具,产品主要销往欧洲和北美),此后成为自由职业者,为客户提供独立的oracle数据库的技术服务和高级性能调整等培训,同时提供ITPUB华南和华东的培训。

目前服务于国内某大型电子商务网站,维护系统数据库并提供开发支持.

擅长的技术领域:oracle

目前的工作动态:alibaba  DBA

个人Blog:http://blog.itpub.net/biti_riany

E-mail:[email protected]

本文地址:http://com.8s8s.com/it/it19466.htm