實戰(zhàn):ORACLE SQL Performance Analyzer的使用

通過 SPA，您可以根據(jù)各種更改類型（如初始化參數(shù)更改、優(yōu)化器統(tǒng)計刷新和數(shù)據(jù)庫升級）播放特定的

SQL 或整個 SQL 負載，然后生成比較報告，幫助您評估它們的影響.

在 Oracle Database 11g 之前的版本中，我必須捕獲所有 SQL 語句，通過跟蹤運行這些語句，

然后得到執(zhí)行計劃 — 這是一項極其耗時又極易出錯的任務(wù)。有了新版本之后，我不需要再那樣做了，

我改用非常簡單而有效的 SQL Performance Analyzer。

---使用場景

1.數(shù)據(jù)庫升級

2.實施優(yōu)化建議

3.更改方案

4.收集統(tǒng)計信息

5.更改數(shù)據(jù)庫參數(shù)

6.更改操作系統(tǒng)和硬件

create tablespace test

datafile 'E:\APP\ADMINISTRATOR\ORADATA\ORCL\test01.DBF'

size 5000m

autoextend on

next 100m maxsize unlimited

extent management local autoallocate

segment   space management auto;

create table t1

(

sid int not null ,

sname varchar2(10)

)

tablespace test;

-2.-循環(huán)導(dǎo)入數(shù)據(jù)

declare

        maxrecords constant int:=1000000;

        i int :=1;

    begin

        for i in 1..maxrecords loop

          insert into t1 values(i,'ocpyang');

        end loop;

    dbms_output.put_line(' 成功錄入數(shù)據(jù)！ ');

    commit;

    end; 

/

update t1 set sname='蘇州' where sid=500001;

update t1 set sname='南京' where sid=600001;

---3.收集統(tǒng)計信息

exec dbms_stats.gather_table_stats(USER,'T1',CASCADE=>TRUE)

alter system flush shared_pool;

---4.執(zhí)行查詢

select count(*) from t1 where sid<=100;

select count(*) from t1 where sid<=500;

select count(*) from t1 where sid>50000;

---5.新建STS

BEGIN

  DBMS_SQLTUNE.DROP_SQLSET(

    sqlset_name => 'OCPYANG_STS'

    );

END;

/

BEGIN

  DBMS_SQLTUNE.CREATE_SQLSET(

    sqlset_name => 'OCPYANG_STS', 

    sqlset_owner => 'SYS',

    description  => 'ocpyangtest');

END;

/

---6.加載sql優(yōu)化集

set serveroutput on

DECLARE 

cur01 dbms_sqltune.sqlset_cursor;

BEGIN 

open cur01 for select value(a) from table(dbms_sqltune.select_cursor_cache

(

basic_filter => 'sql_text like ''%t1%'' and parsing_schema_name =''SYS''',

attribute_list => 'ALL'

)

) a; 

dbms_sqltune.load_sqlset(

sqlset_name => 'OCPYANG_STS',

populate_cursor => cur01);

close cur01;

END; 

/

/*********有兩個參數(shù)值得特別說明：

1）SELECT_CURSOR_CACHE的第一個參數(shù)是basic_filter ，它可以取的值有：

  sql_id                   VARCHAR(13),

  force_matching_signature NUMBER,

  sql_text                 CLOB,

  object_list              sql_objects,

  bind_data                RAW(2000),

  parsing_schema_name      VARCHAR2(30),

  module                   VARCHAR2(48),

  action                   VARCHAR2(32),

  elapsed_time             NUMBER,

  cpu_time                 NUMBER,

  buffer_gets              NUMBER,

  disk_reads               NUMBER,

  direct_writes            NUMBER,

  rows_processed           NUMBER,

  fetches                  NUMBER,

  executions               NUMBER,

  end_of_fetch_count       NUMBER,

  optimizer_cost           NUMBER,

  optimizer_env            RAW(1000),

  priority                 NUMBER,

  command_type             NUMBER,

  first_load_time          VARCHAR2(19),

  stat_period              NUMBER,

  active_stat_period       NUMBER,

  other                    CLOB,

  plan_hash_value          NUMBER,

  sql_plan                 sql_plan_table_type,

  bind_list                sql_binds

2）SELECT_CURSOR_CACHE的最后一個參數(shù)是attribute_list

BASIC (default) -all attributes (such as execution statistics and binds) are returned except the plans The execution context is always part of the result.

TYPICAL - BASIC + SQL plan (without row source statistics) and without object reference list

ALL - return all attributes

Comma separated list of attribute names this allows to return only a subset of SQL attributes: EXECUTION_STATISTICS, BIND_LIST, OBJECT_LIST, SQL_PLAN,SQL_PLAN_STATISTICS: similar to SQL_PLAN + row source statistics

*********/

---7.查詢sql優(yōu)化集

select sql_id,sql_text from dba_sqlset_statements

where sqlset_name='OCPYANG_STS' and sql_text like '% from t1%';

---8.新建SPA

var v_task varchar2(64);

begin

:v_task:=dbms_sqlpa.create_analysis_task(

sqlset_name => 'OCPYANG_STS', 

task_name => 'SPA01'

);

end;

/

/**********語法

Syntax

SQL text format. This form of the function is called to prepare the analysis of a single statement given its text.

DBMS_SQLPA.CREATE_ANALYSIS_TASK(

  sql_text         IN CLOB,

  bind_list        IN sql_binds := NULL,

  parsing_schema   IN VARCHAR2  := NULL,

  task_name        IN VARCHAR2  := NULL,

  description      IN VARCHAR2  := NULL)

RETURN VARCHAR2;

SQL ID format. This form of the function is called to prepare the analysis of a single statement from the cursor cache given its identifier.

DBMS_SQLPA.CREATE_ANALYSIS_TASK(

  sql_id           IN VARCHAR2,

  plan_hash_value  IN NUMBER    := NULL,

  task_name        IN VARCHAR2  := NULL,

  description      IN VARCHAR2  := NULL)

RETURN VARCHAR2;

Workload Repository format. This form of the function is called to prepare the analysis of a single statement from the workload repository given a range of snapshot identifiers.

DBMS_SQLPA.CREATE_ANALYSIS_TASK(

  begin_snap       IN NUMBER,

  end_snap         IN NUMBER,

  sql_id           IN VARCHAR2,

  plan_hash_value  IN NUMBER    := NULL,

  task_name        IN VARCHAR2  := NULL,

  description      IN VARCHAR2  := NULL)

RETURN VARCHAR2;

SQLSET format. This form of the function is called to prepare the analysis of a SQL tuning set.

DBMS_SQLPA.CREATE_ANALYSIS_TASK(

  sqlset_name       IN VARCHAR2,

  basic_filter      IN VARCHAR2 :=  NULL,

  order_by          IN VARCHAR2 :=  NULL,

  top_sql           IN VARCHAR2 :=  NULL,

  task_name         IN VARCHAR2 :=  NULL,

  description       IN VARCHAR2 :=  NULL

  sqlset_owner      IN VARCHAR2 :=  NULL)

RETURN VARCHAR2;

**********/

---9.執(zhí)行SPA

begin

dbms_sqlpa.execute_analysis_task

(

task_name => 'SPA01',

execution_type => 'test execute',

execution_name => 'before_change'

);

end;

/

/*********語法

DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(

   task_name         IN VARCHAR2,

   execution_type    IN VARCHAR2               := 'test execute',

   execution_name    IN VARCHAR2               := NULL,

   execution_params  IN dbms_advisor.argList   := NULL,

   execution_desc    IN VARCHAR2               := NULL)

 RETURN VARCHAR2;

DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(

   task_name         IN VARCHAR2,

   execution_type    IN VARCHAR2               := 'test execute',

   execution_name    IN VARCHAR2               := NULL,

   execution_params  IN dbms_advisor.argList   := NULL,

   execution_desc    IN VARCHAR2               := NULL);

*********/

---10.改變

create index index_01 on t1(sid,sname)

tablespace test;

exec dbms_stats.gather_table_stats(USER,'T1',CASCADE=>TRUE)

---11.改變后執(zhí)行

begin

dbms_sqlpa.execute_analysis_task

(

task_name => 'SPA01',

execution_type => 'test execute',

execution_name => 'after_change'

);

end;

/

col TASK_NAME format a30

col EXECUTION_NAME for a30

select execution_name,

status,

execution_end

from DBA_ADVISOR_EXECUTIONS

where task_name='SPA01'

order by execution_end

/

EXECUTION_NAME                 STATUS      EXECUTION_END

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

before_change                  COMPLETED   2014-05-28 15:43:58

after_change                   COMPLETED   2014-05-28 15:44:58

---12.執(zhí)行任務(wù)比較

begin 

dbms_sqlpa.EXECUTE_ANALYSIS_TASK(

task_name        => 'SPA01', 

execution_type   => 'compare performance', 

execution_params => dbms_advisor.arglist(

'execution_name1', 

'before_change', 

'execution_name2', 

'after_change'));

end;

/

---13.生產(chǎn)報告

set serveroutput on size 999999

set long 100000000

set pagesize 0

set linesize 200

set longchunksize 200

set trimspool on

spool e:\report.txt

select DBMS_SQLPA.REPORT_ANALYSIS_TASK('SPA01') from dual;

spool off;