使用SPA（sql performance Analyzer）
SPA比较在数据库更新前后定义在特定sql优化集中的特定sql语句的性能，重大升级，参数改变，索引或是统计信息的改变。
1.建立测试环境
create table t1 as
select * from dba_objects;
insert into t1 select * from t1;
commit;
在t1上没有任何索引，收集其统计信息：
exec dbms_stats.gather_table_stats(ownname => 'scott',tabname =>'T1',cascade=>true);
查看dba_tables表
select table_name,num_rows from dba_tables where table_name='T1';
TABLE_NAME NUM_ROWS
------------------------------ ----------
T1 145164
统计信息收集完毕。接着清空共享池，接受新的负载（生产库慎用）：
alter system flush shared_pool;
2.执行查询
select count(*) from t1 where object_id=100;
select count(*) from t1 where object_id<100;
select count(*) from t1 where object_id=1000;
select count(*) from t1 where object_id<=1000;
稍后会在object_id列上建立一个索引，并对比建立索引前后sql语句的性能。
3.创建sql优化集
exec dbms_sqltune.create_sqlset(sqlset_name=>'sql_replay_test');
4.加载sql优化集
下面的语句是从cursor_cache中寻找t1表，并将包含该表的sql语句装入sql优化集：
declare
l_cursor dbms_sqltune.sqlset_cursor;
begin
open l_cursor for
select value(a)
from table(dbms_sqltune.select_cursor_cache(basic_filter => 'sql_text like ''%t1%'' and parsing_schema_name=''SCOTT''',
attribute_list => 'ALL')) a;
dbms_sqltune.load_sqlset(sqlset_name => 'sql_replay_test',
populate_cursor => l_cursor);
end;
5.查询sql优化集
select sql_text
from dba_sqlset_statements
where sqlset_name='sql_replay_test';
SQL_TEXT
--------------------------------------------------------------------------------
SELECT VALUE(A) FROM TABLE(DBMS_SQLTUNE.SELECT_CURSOR_CACHE(BASIC_FILTER => 'sql
select count(*) from t1 where object_id<=1000
select count(*) from t1 where object_id<100
SELECT /*+ first_rows(1) */ sql_id, force_matching_signature, sql_text, cast(NU
select count(*) from t1 where object_id=100
select count(*) from t1 where object_id=1000
6.输出sql优化集
variable v_task varchar2(64);
exec :v_task:=dbms_sqlpa.create_analysis_task(sqlset_name=>'sql_replay_test');
print :v_task
v_task
---------
TASK_281
7.在分析之前执行
在数据库改动之前执行优化集的内容，收集性能信息：
begin
dbms_sqlpa.execute_analysis_task(task_name => :v_task,
execution_type => 'test execute',
execution_name => 'before_change');
end;
8.做出必要的改动
在t1表的object_id列加一个索引，来提高性能，并重新收集统计信息。
create index t1_index_id on t1(object_id);
exec dbms_stats.gather_table_stats(ownname => 'scott',tabname =>'T1',cascade=>true);
9.创建索引后执行分析任务
print :v_task
v_task
---------
TASK_281
begin
dbms_sqlpa.execute_analysis_task(task_name => :v_task,
execution_type => 'test execute',
execution_name => 'after_change');
end;
10.执行分析任务比较
print :v_task
v_task
---------
TASK_281
begin
dbms_sqlpa.execute_analysis_task(task_name => :v_task,
execution_type => 'compare performance',
execution_params => dbms_advisor.argList('execution_name1',
'before_change',
'execution_name2',
'after_change'));
end;
11.输出最终分析结果
select dbms_sqlpa.report_analysis_task(:v_task) from dual;
General Information
--------------------------------------------------------------------------------
Task Information: Workload Information:
--------------------------------------------- --------------------------------
Task Name : TASK_281 SQL Tuning Set Name : sq
Task Owner : SCOTT SQL Tuning Set Owner : SC
Description : Total SQL Statement Count : 6
Execution Information:
--------------------------------------------------------------------------------
Execution Name : EXEC_264 Started : 04/20/2016 08:0
Execution Type : COMPARE PERFORMANCE Last Updated : 04/20/2016 08:0
Description : Global Time Limit : UNLIMITED
Scope : COMPREHENSIVE Per-SQL Time Limit : UNUSED
Status : COMPLETED Number of Errors : 0
Analysis Information:
--------------------------------------------------------------------------------
Before Change Execution: After Change Execution:
--------------------------------------------- --------------------------------
Execution Name : before_change Execution Name : after_cha
Execution Type : TEST EXECUTE Execution Type : TEST EXEC
Scope : COMPREHENSIVE Scope : COMPREHEN
Status : COMPLETED Status : COMPLETED
Started : 04/20/2016 07:53:20 Started : 04/20/201
Last Updated : 04/20/2016 07:53:22 Last Updated : 04/20/201
Global Time Limit : UNLIMITED Global Time Limit : UNLIMITED
Per-SQL Time Limit : UNUSED Per-SQL Time Limit : UNUSED
Number of Errors : 0 Number of Errors : 0
---------------------------------------------
Comparison Metric: ELAPSED_TIME
------------------
Workload Impact Threshold: 1%
--------------------------
SQL Impact Threshold: 1%
----------------------
Report Summary
--------------------------------------------------------------------------------
Projected Workload Change Impact:
-------------------------------------------
Overall Impact : 17.33%
Improvement Impact : 42.05%
Regression Impact : -24.72%
SQL Statement Count
-------------------------------------------
SQL Category SQL Count Plan Change Count
Overall 6 4
Improved 4 4
Regressed 2 0
Top 6 SQL Sorted by Absolute Value of Change Impact on the Workload
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
| | | Impact on | Execution | Metric | Metric | Impact
| object_id | sql_id | Workload | Frequency | Before | After | on SQL
--------------------------------------------------------------------------------
| 16 | 04uqmkx8ra2p6 | -15.84% | 1 | 18185 | 27729 | -52.48%
| 18 | 7p08u8zjtkayn | 11.76% | 1 | 7118 | 33 | 99.54%
| 17 | 2575930rsg6qb | 11.2% | 1 | 6893 | 149 | 97.84%
| 20 | 8gj6r733n51jn | 9.71% | 1 | 5875 | 27 | 99.54%
| 21 | csrkbp9cvp4g4 | 9.38% | 1 | 5683 | 31 | 99.45%
| 19 | 84fx5r1d8n00p | -8.88% | 1 | 16484 | 21831 | -32.44%
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
v_task
---------
TASK_281
这里的总体影响是17.33%，提高了。