6 Comparing SQL Trials

6.1.1 Analyzing SQL Performance Using Oracle Enterprise Manager

This section describes how to analyze SQL performance before and after the system change using Oracle Enterprise Manager.

To analyze SQL performance using Enterprise Manager:

1. On the Guided Workflow page, click the Execute icon for Compare Step 2 and Step 3.

   The Run SQL Trial Comparison page appears.

   
   Description of the illustration spa_run_trial_compare.gif
   
   

   In this example, the SQL_TRIAL_1241213421833 and SQL_TRIAL_1241213881923 trials are selected for comparison.

2. To compare trials other than those listed by default, select the desired trials in the Trial 1 Name and Trial 2 Name lists.

   Note that you cannot compare a statistical trial with a trial that tests the explain plan only.

3. In the Comparison Metric list, select the comparison metric to use for the comparison analysis:

   • Elapsed Time

   • CPU Time

   • User I/O Time

   • Buffer Gets

   • Physical I/O

   • Optimizer Cost

   • I/O Interconnect Bytes

   Optimizer Cost is the only comparison metric available if you generated execution plans only in the SQL trials.

   To perform the comparison analysis by using more than one comparison metric, perform separate comparison analyses by repeating this procedure with different metrics.

4. In the Schedule section:

   1. In the Time Zone list, select your time zone code.

   2. Select Immediately to start the task now, or Later to schedule the task to start at a time specified using the Date and Time fields.

5. Click Submit.

   The Guided Workflow page appears when the comparison analysis begins.

   The status icon of this step changes to an arrow icon while the comparison analysis is in progress. To refresh the status icon, click Refresh. Depending on the amount of performance data collected from the pre-change and post-change executions, the comparison analysis may take a long time to complete. After the comparison analysis is completed, the Status icon changes to a check mark and the Execute icon for the next step is enabled.

6. Once SQL Performance Analyzer has analyzed the pre-change and post-change performance data, generate a SQL Performance Analyzer report that you can use for further analysis.

   On the Guided Workflow page, click the Execute icon for View Trial Comparison Report.

   The SQL Performance Analyzer Task Report page appears. Review the report, as described in "Reviewing the SQL Performance Analyzer Report Using Oracle Enterprise Manager".

6.1.2 Reviewing the SQL Performance Analyzer Report Using Oracle Enterprise Manager

When a SQL Performance Analyzer task is completed, the resulting data is generated into a SQL Performance Analyzer report that compares the pre-change and post-change SQL performance.

Figure 6-1 shows a sample SQL Performance Analyzer report. This sample report uses the elapsed time comparison metric to compare the pre-change and post-change executions of a SQL workload.

Figure 6-1 SQL Performance Analyzer Report

Description of "Figure 6-1 SQL Performance Analyzer Report"

To generate and review the SQL Performance Analyzer report:

1. On the Software and Support page, under Real Application Testing, click SQL Performance Analyzer.

   The SQL Performance Analyzer page appears. A list of existing SQL Performance Analyzer tasks are displayed.

2. Under SQL Performance Analyzer Tasks, select the task for which you want to view a SQL Performance Analyzer report and click View Latest Report.

   The SQL Performance Analyzer Task Report page appears.

3. Review the general information about the performance analysis, as described in "Reviewing the SQL Performance Analyzer Report: General Information".

4. Review general statistics, as described in "Reviewing the SQL Performance Analyzer Report: Global Statistics".

5. Optionally, review the detailed statistics, as described in "Reviewing the SQL Performance Analyzer Report: Global Statistics Details".

6. To generate an active report, click Save to generate and save the report, or Mail to generate and mail the report as an HTML attachment.

   Active reports include information about the top SQL statements from each category (such as improved, regressed, and changed plans) with pre-change and post-change statistics, explain plans, and task summary.

   For more information, see "About SQL Performance Analyzer Active Reports".

6.1.3 Tuning Regressed SQL Statements Using Oracle Enterprise Manager

After reviewing the SQL Performance Analyzer report, you should tune any regressed SQL statements that are identified after comparing the SQL performance. If there are large numbers of SQL statements that appear to have regressed, you should try to identify the root cause and make system-level changes to rectify the problem. In cases when only a few SQL statements have regressed, consider using one of the following tuning methods to implement a point solution for them:

• Creating SQL Plan Baselines

• Running SQL Tuning Advisor

After tuning the regressed SQL statements, you should test these changes using SQL Performance Analyzer. Run a new SQL trial on the test system, followed by a second comparison (between this new SQL trial and the first SQL trial) to validate your results. Once SQL Performance Analyzer shows that performance has stabilized, the testing is complete. Implement the fixes from this step to your production system.

Starting with Oracle Database 11g Release 1, SQL Tuning Advisor performs an alternative plan analysis when tuning a SQL statement. SQL Tuning Advisor searches the current system for previous execution plans, including the plans from the first SQL trial. If the execution plans from the first SQL trial differ from those of the second SQL trial, SQL Tuning Advisor will recommend the plans from the first SQL trial. If these execution plans produce better performance, you can create plan baselines using the plans from the first SQL trial.

6.2.1 Analyzing SQL Performance Using APIs

After the post-change SQL performance data is built, you can compare the pre-change version of performance data to the post-change version. Run a comparison analysis using the DBMS_SQLPA.EXECUTE_ANALYSIS_TASK procedure or function.

To compare the pre-change and post-change SQL performance data:

1. Call the EXECUTE_ANALYSIS_TASK procedure or function using the following parameters:

   • Set the task_name parameter to the name of the SQL Performance Analyzer task.

   • Set the execution_type parameter to COMPARE PERFORMANCE. This setting will analyze and compare two versions of SQL performance data.

   • Specify a name to identify the execution using the execution_name parameter. If not specified, it will be generated by SQL Performance Analyzer and returned by the function.

   • Specify two versions of SQL performance data using the execution_params parameters. The execution_params parameters are specified as (name, value) pairs for the specified execution. Set the execution parameters that are related to comparing and analyzing SQL performance data as follows:

     • Set the execution_name1 parameter to the name of the first execution (before the system change was made). This value should correspond to the value of the execution_name parameter specified in "Creating a Pre-Change SQL Trial Using APIs".

     • Set the execution_name2 parameter to the name of the second execution (after the system change was made). This value should correspond to the value of the execution_name parameter specified in "Creating a Post-Change SQL Trial Using APIs" when you executed the SQL workload after the system change. If the caller does not specify the executions, then by default SQL Performance Analyzer will always compare the last two task executions.

     • Set the comparison_metric parameter to specify an expression of execution statistics to use in the performance impact analysis. Possible values include the following metrics or any combination of them: elapsed_time (default), cpu_time, buffer_gets, disk_reads, direct_writes, optimizer_cost, and io_interconnect_bytes.

     For other possible parameters that you can set for comparison, see the description of the DBMS_SQLPA package in Oracle Database PL/SQL Packages and Types Reference.

   The following example illustrates a function call:

   EXEC DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(task_name => 'my_spa_task', - 
          execution_type => 'COMPARE PERFORMANCE', -
          execution_name => 'my_exec_compare', -
          execution_params => dbms_advisor.arglist(-
                                'comparison_metric', 'buffer_gets'));
   

2. Call the REPORT_ANALYSIS_TASK function using the following parameters:

   • Set the task_name parameter to the name of the SQL Performance Analyzer task.

   • Set the execution_name parameter to the name of the execution to use. This value should match the execution_name parameter of the execution for which you want to generate a report.

     To generate a report to display the results of:

     • Execution plans generated for the SQL workload, set this value to match the execution_name parameter of the desired EXPLAIN PLAN execution.

     • Execution plans and execution statistics generated for the SQL workload, set this parameter to match the value of the execution_name parameter used in the desired TEST EXECUTE execution.

     • A comparison analysis, set this value to match the execution_name parameter of the desired ANALYZE PERFORMANCE execution.

     If unspecified, SQL Performance Analyzer generates a report for the last execution.

   • Set the type parameter to specify the type of report to generate. Possible values include TEXT (default), HTML, XML, and ACTIVE.

     Active reports provides in-depth reporting using an interactive user interface that enables you to perform detailed analysis even when disconnected from the database or Oracle Enterprise Manager. It is recommended that you use active reports instead of HTML or text reports when possible.

     For information about active reports, see "About SQL Performance Analyzer Active Reports".

   • Set the level parameter to specify the format of the recommendations. Possible values include TYPICAL (default), ALL, BASIC, CHANGED, CHANGED_PLANS, ERRORS, IMPROVED, REGRESSED, TIMEOUT, UNCHANGED, UNCHANGED_PLANS, and UNSUPPORTED.

   • Set the section parameter to specify a particular section to generate in the report. Possible values include SUMMARY (default) and ALL.

   • Set the top_sql parameter to specify the number of SQL statements in a SQL tuning set to generate in the report. By default, the report shows the top 100 SQL statements impacted by the system change.

   To generate an active report, run the following script:

   set trimspool on
   set trim on
   set pages 0
   set linesize 1000
   set long 1000000
   set longchunksize 1000000
   spool spa_active.html
   SELECT DBMS_SQLPA.REPORT_ANALYSIS_TASK(task_name => 'my_spa_task',
             type => 'active', section => 'all') FROM dual;
   spool off
   

   The following example illustrates a portion of a SQL script that you could use to create and display a comparison summary report in text format:

   VAR rep   CLOB;
   EXEC :rep := DBMS_SQLPA.REPORT_ANALYSIS_TASK('my_spa_task', -
                   'text', 'typical', 'summary');
   SET LONG 100000 LONGCHUNKSIZE 100000 LINESIZE 130
   PRINT :rep
   

3. Review the SQL Performance Analyzer report, as described in "Reviewing the SQL Performance Analyzer Report Using APIs".

6.2.2 Reviewing the SQL Performance Analyzer Report Using APIs

The SQL Performance Analyzer report is divided into the following sections:

• General Information

• Result Summary

• Result Details

This section uses a sample report to illustrate how to review the SQL Performance Analyzer report. The sample report uses buffer_gets as the comparison metric to compare the pre-change and post-change executions of a SQL workload.

---------------------------------------------------------------------------------------------
General Information
---------------------------------------------------------------------------------------------

 Task Information:                              Workload Information:
 ---------------------------------------------  ---------------------------------------------
  Task Name    : my_spa_task                     SQL Tuning Set Name        : my_sts
  Task Owner   : APPS                            SQL Tuning Set Owner       : APPS
  Description  :                                 Total SQL Statement Count  : 101

Execution Information:
---------------------------------------------------------------------------------------------
  Execution Name  : my_exec_compare        Started             : 05/21/2007 11:30:09
  Execution Type  : ANALYZE PERFORMANCE    Last Updated        : 05/21/2007 11:30:10
  Description     :                        Global Time Limit   : UNLIMITED
  Scope           : COMPREHENSIVE          Per-SQL Time Limit  : UNUSED
  Status          : COMPLETED              Number of Errors    : 0

Analysis Information:
---------------------------------------------------------------------------------------------
 Comparison Metric: BUFFER_GETS
 ------------------
 Workload Impact Threshold: 1%
 --------------------------
 SQL Impact Threshold: 1%
 ----------------------
 Before Change Execution:                       After Change Execution:
 ---------------------------------------------  ---------------------------------------------
  Execution Name      : my_exec_BEFORE_change    Execution Name      : my_exec_AFTER_change
  Execution Type      : TEST EXECUTE             Execution Type      : TEST EXECUTE
  Description         :                          Description         :
  Scope               : COMPREHENSIVE            Scope               : COMPREHENSIVE
  Status              : COMPLETED                Status              : COMPLETED
  Started             : 05/21/2007 11:22:06      Started             : 05/21/2007 11:25:56
  Last Updated        : 05/21/2007 11:24:01      Last Updated        : 05/21/2007 11:28:30
  Global Time Limit   : 1800                     Global Time Limit   : 1800
  Per-SQL Time Limit  : UNUSED                   Per-SQL Time Limit  : UNUSED
  Number of Errors    : 0                        Number of Errors    : 0

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

Report Summary
---------------------------------------------------------------------------------------------
 
Projected Workload Change Impact:
-------------------------------------------
 Overall Impact      :   47.94%
 Improvement Impact  :   58.02%
 Regression Impact   :  -10.08%
 
SQL Statement Count
-------------------------------------------
 SQL Category  SQL Count  Plan Change Count
 Overall       101        6
 Improved      2          2
 Regressed     1          1
 Unchanged     98         3
.
.
.
---------------------------------------------------------------------------------------------

SQL Statements Sorted by their Absolute Value of Change Impact on the Workload
---------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------
|           |               | Impact on | Execution |  Metric  |  Metric  | Impact | Plan   |
| object_id | sql_id        | Workload  | Frequency |  Before  |  After   | on SQL | Change |
---------------------------------------------------------------------------------------------
|    205    | 73s2sgy2svfrw |   29.01%  |   100000  |  1681683 |  220590  | 86.88% |   y    |
|    206    | gq2a407mv2hsy |   29.01%  |   949141  |  1681683 |  220590  | 86.88% |   y    |
|    204    | 2wtgxbjz6u2by |  -10.08%  |   478254  |  1653012 |  2160529 | -30.7% |   y    |
---------------------------------------------------------------------------------------------

----------------------------------------------------------------------------------
                             SQL STATEMENTS WITH ERRORS
----------------------------------------------------------------------------------
SQL ID        Error
------------- --------------------------------------------------------------------
47bjmcdtw6htn ORA-00942: table or view does not exist
br61bjp4tnf7y ORA-00920: invalid relational operator
----------------------------------------------------------------------------------

SQL Details:
-----------------------------
 Object ID            : 204
 Schema Name          : APPS
 SQL ID               : 2wtgxbjz6u2by
 Execution Frequency  : 1
 SQL Text             : SELECT /* my_query_14_scott */ /*+ ORDERED INDEX(t1)
                        USE_HASH(t1) */ 'B' || t2.pg_featurevalue_05_id
                        pg_featurevalue_05_id, 'r' || t4.elementrange_id
                        pg_featurevalue_15_id, 'G' || t5.elementgroup_id
                        pg_featurevalue_01_id, 'r' || t6.elementrange_id . . .
.
.
.
---------------------------------------------------------------------------------------------

Execution Statistics:
-----------------------------
---------------------------------------------------------------------------------------
|              | Impact on | Value   | Value   | Impact     | % Workload | % Workload |
| Stat Name    | Workload  | Before  | After   | on SQL     | Before     | After      |
---------------------------------------------------------------------------------------
| elapsed_time |    -95.54%|  36.484 | 143.161 |   -292.39% |     32.68% |     94.73% |
| parse_time   |    -12.37%|    .004 |    .062 |     -1450% |       .85% |     11.79% |
| exec_elapsed |    -95.89%|   36.48 | 143.099 |   -292.27% |     32.81% |     95.02% |
| exec_cpu     |    -19.73%|  36.467 |  58.345 |    -59.99% |     32.89% |     88.58% |
| buffer_gets  |    -10.08%| 1653012 | 2160529 |     -30.7% |     32.82% |     82.48% |
| cost         |     12.17%|   11224 |    2771 |     75.31% |     16.16% |      4.66% |
| reads        |  -1825.72%|    4091 |  455280 | -11028.82% |     16.55% |     96.66% |
| writes       |     -1500%|       0 |      15 |     -1500% |         0% |       100% |
| rows         |           |     135 |     135 |            |            |            |
---------------------------------------------------------------------------------------

Notes:
-----------------------------
Before Change:
1. The statement was first executed to warm the buffer cache.
2. Statistics shown were averaged over next 9 executions.

After Change:
1. The statement was first executed to warm the buffer cache.
2. Statistics shown were averaged over next 9 executions.

Findings (2):
-----------------------------
1. The performance of this SQL has regressed.
2. The structure of the SQL execution plan has changed.
---------------------------------------------------------------------------------------------

Execution Plan Before Change:
-----------------------------
 Plan Id          : 1
 Plan Hash Value  : 3412943215
 
----------------------------------------------------------------------------------------------------------
| Id   | Operation                           | Name                | Rows   | Bytes   | Cost  | Time     |
----------------------------------------------------------------------------------------------------------
|    0 | SELECT STATEMENT                    |                     |      1 |     126 | 11224 | 00:02:15 |
|    1 |   HASH GROUP BY                     |                     |      1 |     126 | 11224 | 00:02:15 |
|    2 |    NESTED LOOPS                     |                     |      1 |     126 | 11223 | 00:02:15 |
|  * 3 |     HASH JOIN                       |                     |      1 |     111 | 11175 | 00:02:15 |
|  * 4 |      TABLE ACCESS FULL              | LU_ELEMENTGROUP_REL |      1 |      11 |   162 | 00:00:02 |
|  * 5 |      HASH JOIN                      |                     |    487 |   48700 | 11012 | 00:02:13 |
|    6 |       MERGE JOIN                    |                     |     14 |     924 |  1068 | 00:00:13 |
|    7 |        SORT JOIN                    |                     |   5391 |  274941 |  1033 | 00:00:13 |
|  * 8 |         HASH JOIN                   |                     |   5391 |  274941 |   904 | 00:00:11 |
|  * 9 |          TABLE ACCESS FULL          | LU_ELEMENTGROUP_REL |    123 |    1353 |   175 | 00:00:03 |
| * 10 |          HASH JOIN                  |                     |   5352 |  214080 |   729 | 00:00:09 |
| * 11 |           TABLE ACCESS FULL         | LU_ITEM_293         |   5355 |  128520 |    56 | 00:00:01 |
| * 12 |           TABLE ACCESS FULL         | ADM_PG_FEATUREVALUE |   1629 |   26064 |   649 | 00:00:08 |
| * 13 |        FILTER                       |                     |        |         |       |          |
| * 14 |         SORT JOIN                   |                     |      1 |      15 |    36 | 00:00:01 |
| * 15 |          TABLE ACCESS FULL          | LU_ELEMENTRANGE_REL |      1 |      15 |    35 | 00:00:01 |
|   16 |       INLIST ITERATOR               |                     |        |         |       |          |
| * 17 |        TABLE ACCESS BY INDEX ROWID  | FACT_PD_OUT_ITM_293 | 191837 | 6522458 |  9927 | 00:02:00 |
|   18 |         BITMAP CONVERSION TO ROWIDS |                     |        |         |       |          |
| * 19 |          BITMAP INDEX SINGLE VALUE  | FACT_274_PER_IDX    |        |         |       |          |
| * 20 |     TABLE ACCESS FULL               | LU_ELEMENTRANGE_REL |      1 |      15 |    49 | 00:00:01 |
----------------------------------------------------------------------------------------------------------
.
.
.

Execution Plan After Change:
-----------------------------
 Plan Id          : 102
 Plan Hash Value  : 1923145679
 
------------------------------------------------------------------------------------------------------
| Id   | Operation                           | Name                | Rows | Bytes  | Cost | Time     |
------------------------------------------------------------------------------------------------------
|    0 | SELECT STATEMENT                    |                     |    1 |    126 | 2771 | 00:00:34 |
|    1 |   HASH GROUP BY                     |                     |    1 |    126 | 2771 | 00:00:34 |
|    2 |    NESTED LOOPS                     |                     |    1 |    126 | 2770 | 00:00:34 |
|  * 3 |     HASH JOIN                       |                     |    1 |    111 | 2722 | 00:00:33 |
|  * 4 |      HASH JOIN                      |                     |    1 |    100 | 2547 | 00:00:31 |
|  * 5 |       TABLE ACCESS FULL             | LU_ELEMENTGROUP_REL |    1 |     11 |  162 | 00:00:02 |
|    6 |       NESTED LOOPS                  |                     |      |        |      |          |
|    7 |        NESTED LOOPS                 |                     |  484 |  43076 | 2384 | 00:00:29 |
|  * 8 |         HASH JOIN                   |                     |   14 |    770 |  741 | 00:00:09 |
|    9 |          NESTED LOOPS               |                     |    4 |    124 |  683 | 00:00:09 |
| * 10 |           TABLE ACCESS FULL         | LU_ELEMENTRANGE_REL |    1 |     15 |   35 | 00:00:01 |
| * 11 |           TABLE ACCESS FULL         | ADM_PG_FEATUREVALUE |    4 |     64 |  649 | 00:00:08 |
| * 12 |          TABLE ACCESS FULL          | LU_ITEM_293         | 5355 | 128520 |   56 | 00:00:01 |
|   13 |         BITMAP CONVERSION TO ROWIDS |                     |      |        |      |          |
| * 14 |          BITMAP INDEX SINGLE VALUE  | FACT_274_ITEM_IDX   |      |        |      |          |
| * 15 |        TABLE ACCESS BY INDEX ROWID  | FACT_PD_OUT_ITM_293 |   36 |   1224 | 2384 | 00:00:29 |
| * 16 |      TABLE ACCESS FULL              | LU_ELEMENTGROUP_REL |  123 |   1353 |  175 | 00:00:03 |
| * 17 |     TABLE ACCESS FULL               | LU_ELEMENTRANGE_REL |    1 |     15 |   49 | 00:00:01 |
------------------------------------------------------------------------------------------------------

6.2.3 Comparing SQL Tuning Sets Using APIs

You can compare two SQL tuning sets using SQL Performance Analyzer APIs. For example, while using Database Replay, you may have captured a SQL tuning set on the production system during workload capture, and another SQL tuning set on a test system during workload replay. You can then use SQL Performance Analyzer to compare these SQL tuning sets, without having to re-execute the SQL statements. This is useful in cases where you already have another utility to run your workload before and after making the system change, such as a custom script.

When comparing SQL tuning sets, SQL Performance Analyzer uses the runtime statistics captured in the SQL tuning sets to perform its comparison analysis, and reports on any new or missing SQL statements that are found in one SQL tuning set, but not in the other. Any changes in execution plans between the two SQL tuning sets are also reported. For each SQL statement in both SQL tuning sets, improvement and regression findings are reported for each SQL statement—calculated based on the average statistic value per execution—and for the entire workload—calculated based on the cumulative statistic value.

To compare SQL tuning sets using the DBMS_SQLPA package:

1. Create a SQL Performance Analyzer task:

   VAR aname varchar2(30);
   EXEC :aname := 'compare_s2s';
   EXEC :aname := DBMS_SQLPA.CREATE_ANALYSIS_TASK(task_name => :aname);
   

   It is not necessary to associate a SQL tuning set to the task during creation.

2. Create the first SQL trial and convert the first SQL tuning set:

   EXEC DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(task_name => :aname, -
             execution_type => 'convert sqlset', -
             execution_name => 'first trial', -
             execution_params => DBMS_ADVISOR.ARGLIST(
                                  'sqlset_name', 'my_first_sts', -
                                  'sqlset_owner', 'APPS'));
   

   Specify the name and owner of the SQL tuning set using the SQLSET_NAME and SQLSET_OWNER task parameters. The content of the SQL tuning set will not be duplicated by the SQL Performance Analyzer task. Instead, a reference to the SQL tuning set is recorded in association to the new SQL trial, which in this example is "first trial".

3. Create a second SQL trial and associate it to the second SQL tuning second to which you want to compare:

   EXEC DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(task_name => :aname, -
             execution_type => 'convert sqlset', -
             execution_name => 'second trial', -
             execution_params => DBMS_ADVISOR.ARGLIST(
                                  'sqlset_name', 'my_second_sts', -
                                  'sqlset_owner', 'APPS'));
   

4. Compare the performance data from the two SQL trials (or SQL tuning sets) by running a comparison analysis:

   EXEC DBMS_SQLPA.EXECUTE_ANALYSIS_TASK(task_name => :aname, -
             execution_type => 'compare', -
             execution_name => 'comparison', -
             execution_params => DBMS_ADVISOR.ARGLIST(
                                  'workload_impact_threshold', 0, -
                                  'sql_impact_threshold', 0));
   

   In this example, the workload and per-SQL impact threshold are set to 0% for comparison (the default value is 1%).

5. After the comparison analysis is complete, generate a SQL Performance Analyzer report using the DBMS_SQLPA.REPORT_ANALYSIS_TASK function.

   For information about generating a SQL Performance Analyzer report using APIs, see "Analyzing SQL Performance Using APIs".

Once the report is generated, review it to identify any differences between the contents of the two SQL tuning sets. Example 6-8 shows the Analysis Information and Report Summary sections of a sample report generated by comparing two SQL tuning sets:

Analysis Information:
------------------------------------------------------------------------------------------------
 Before Change Execution:                          After Change Execution:
 ---------------------------------------------     ---------------------------------------------
  Execution Name      : first trial                 Execution Name      : second trial
  Execution Type      : CONVERT SQLSET              Execution Type      : CONVERT SQLSET
  Status              : COMPLETED                   Status              : COMPLETED
  Started             : …
  Last Updated        : …
 
 Before Change Workload:                           After Change Workload:
 ---------------------------------------------     ---------------------------------------------
  SQL Tuning Set Name        : my_first_sts         SQL Tuning Set Name        : my_second_sts
  SQL Tuning Set Owner       : APPS                 SQL Tuning Set Owner       : APPS
  Total SQL Statement Count  : 5                    Total SQL Statement Count  : 6
 
------------------------------------------------------------------------------------------------
Report Summary
------------------------------------------------------------------------------------------------
 
Projected Workload Change Impact:
-------------------------------------------
 Overall Impact      :  72.32%
 Improvement Impact  :  47.72%
 Regression Impact   :   -.02%
 Missing-SQL Impact  :   33.1%
 New-SQL Impact      :  -8.48%
 
SQL Statement Count
-------------------------------------------
 SQL Category   SQL Count  Plan Change Count
 Overall                7                  1
 Common                 4                  1
  Improved              3                  1
  Regressed             1                  0
 Different              3                  0
  Missing SQL           1                  0
  New SQL               2                  0

As shown in Example 6-8, this report contains two additional categories that are not found in standard SQL Performance Analyzer reports; both categories are grouped under the heading Different:

• Missing SQL

  This category represents all SQL statements that are present in the first SQL tuning set, but are not found in the second SQL tuning set. In this example, only one SQL statement is missing. As shown in Example 6-9, this SQL statement has:

  • A sql_id value of gv7xb8tyd1v91

  • A performance impact on the workload of 33.1% based on the change

  • No performance impact on the SQL statement based on the change because its "Total Metric After" change value is missing

• New SQL

  This category represents all SQL statements that are present in the second SQL tuning set, but are not found in the first SQL tuning set. In this example, only two SQL statements are new in the second SQL tuning set. As shown in Example 6-9, these SQL statements have:

  • sql_id values of 4c8nrqxhtb2sf and 9utadgu5udmh4

  • A total performance impact on the workload of -8.48%

  • Missing "Total Metric Before" change values

Example 6-9 shows a table in the sample report that lists the missing and new SQL statements, as well as other top SQL statements as determined by their impact on the workload:

Top 7 SQL Sorted by Absolute Value of Change Impact on the Workload
------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------
|           |               | Impact on | Total Metric | Total Metric | Impact  | Plan   |
| object_id | sql_id        | Workload  |    Before    |    After     | on SQL  | Change |
------------------------------------------------------------------------------------------
|         4 | 7gj3w9ya4d9sj |    41.04% |       812791 |        36974 |     95% | y      |
|         7 | gv7xb8tyd1v91 |     33.1% |       625582 |              |         | n      |
|         2 | 4c8nrqxhtb2sf |    -8.35% |              |       157782 |         | n      |
|         1 | 22u3tvrt0yr6g |     4.58% |       302190 |       215681 |  28.63% | n      |
|         6 | fgdd0fd56qmt0 |      2.1% |       146128 |       106369 |  27.21% | n      |
|         5 | 9utadgu5udmh4 |     -.13% |              |         2452 |         | n      |
|         3 | 4dtv43awxnmv3 |     -.02% |         3520 |         3890 | -47.35% | n      |
------------------------------------------------------------------------------------------

Once you have identified a SQL statement of interest, you can generate a report for the SQL statement to perform more detailed investigation. For example, you may want to investigate the SQL statement with the sql_id value of 7gj3w9ya4d9sj and object_id value of 4 because it has the highest impact on the workload:

SELECT DBMS_SQLPA.REPORT_ANALYSIS_TASK(task_name => :aname, object_id => 4) rep
FROM dual;

Example 6-10 shows a sample report generated for this SQL statement:

SQL Details:
-----------------------------
 Object ID  : 4
 SQL ID     : 7gj3w9ya4d9sj
 SQL Text   : /* my_csts_query1 */ select * FROM emp where empno=2
 
SQL Execution Statistics (average):
---------------------------------------
---------------------------------------------------------
|              | Impact on | Value   | Value   | Impact |
| Stat Name    | Workload  | Before  | After   | on SQL |
---------------------------------------------------------
| elapsed_time |    41.04% | .036945 | .001849 |    95% |
| cpu_time     |    13.74% | .004772 |  .00185 | 61.24% |
| buffer_gets  |     9.59% |       8 |       2 | 69.01% |
| cost         |    11.76% |       1 |       1 |    10% |
| reads        |     4.08% |       0 |       0 | 63.33% |
| writes       |        0% |       0 |       0 |     0% |
| rows         |           |       0 |       0 |        |
| executions   |           |      22 |      20 |        |
| plan_count   |           |       3 |       2 |        |
---------------------------------------------------------
Findings (2):
-----------------------------
 1. The performance of this SQL has improved.
 2. The structure of the SQL execution plan has changed.
 
Plan Execution Statistics (average):
---------------------------------------
----------------------------------------------------------------------------------
| Statistic Name  | Plans Before Change              | Plans After Change        |
----------------------------------------------------------------------------------
| plan hash value | 440231712  571903972  3634526668 | 571903972  3634526668     |
| --------------- | ---------  ---------  ---------- | ---------  ----------     |
|  schema name    | APPS1      APPS2      APPS2      | APPS2      APPS2          |
|  executions     | 7          5          10         | 10         10             |
|  cost           | 2          1          2          | 1          2              |
|  elapsed_time   | .108429    .000937    .00491     | .000503    .003195        |
|  cpu_time       | .00957     .0012      .0032      | .0005      .0032          |
|  buffer_gets    | 18         0          5          | 0          5              |
|  reads          | 0          0          0          | 0          0              |
|  writes         | 0          0          0          | 0          0              |
|  rows           | 0          0          0          | 0          0              |
----------------------------------------------------------------------------------
Execution Plans Before Change:
-----------------------------
Plan Hash Value  : 440231712
---------------------------------------------------------------------------
| Id | Operation              | Name     | Rows | Bytes | Cost | Time     |
---------------------------------------------------------------------------
|  0 | SELECT STATEMENT       |          |      |       |    2 |          |
|  1 |   PX COORDINATOR       |          |      |       |      |          |
|  2 |    PX SEND QC (RANDOM) | :TQ10000 |    1 |    87 |    2 | 00:00:01 |
|  3 |     PX BLOCK ITERATOR  |          |    1 |    87 |    2 | 00:00:01 |
|  4 |      TABLE ACCESS FULL | EMP      |    1 |    87 |    2 | 00:00:01 |
---------------------------------------------------------------------------
Note
-----
- dynamic sampling used for this statement
Plan Hash Value  : 571903972
----------------------------------------------------------------------------------
| Id | Operation                     | Name       | Rows | Bytes | Cost | Time   |
----------------------------------------------------------------------------------
|  0 | SELECT STATEMENT              |            |      |       |    1 |        |
|  1 |   TABLE ACCESS BY INDEX ROWID | EMP        |    1 |    87 |    1 |00:00:01|
|  2 |    INDEX UNIQUE SCAN          | MY_EMP_IDX |    1 |       |    0 |        |
----------------------------------------------------------------------------------
Plan Hash Value  : 3634526668
--------------------------------------------------------------------
| Id | Operation           | Name | Rows | Bytes | Cost | Time     |
--------------------------------------------------------------------
|  0 | SELECT STATEMENT    |      |      |       |    2 |          |
|  1 |   TABLE ACCESS FULL | EMP  |    1 |    87 |    2 | 00:00:01 |
--------------------------------------------------------------------
Note
-----
- dynamic sampling used for this statement
 
Executions Plan After Change:
-----------------------------
Plan Hash Value  : 571903972
----------------------------------------------------------------------------------
| Id | Operation                     | Name       | Rows | Bytes | Cost | Time   |
----------------------------------------------------------------------------------
|  0 | SELECT STATEMENT              |            |      |       |    1 |        |
|  1 |   TABLE ACCESS BY INDEX ROWID | EMP        |    1 |    87 |    1 |00:00:01|
|  2 |    INDEX UNIQUE SCAN          | MY_EMP_IDX |    1 |       |    0 |        |
----------------------------------------------------------------------------------
Plan Hash Value  : 3634526668
--------------------------------------------------------------------
| Id | Operation           | Name | Rows | Bytes | Cost | Time     |
--------------------------------------------------------------------
|  0 | SELECT STATEMENT    |      |      |       |    2 |          |
|  1 |   TABLE ACCESS FULL | EMP  |    1 |    87 |    2 | 00:00:01 |
--------------------------------------------------------------------
Note
-----
- dynamic sampling used for this statement
----------------------------------------------------------------------------------

The SQL Execution Statistics section shows the average runtime statistics (per execution) of the SQL statement. The data in this table reveals that this SQL statement is present in both SQL tuning sets, but that it has only three execution plans in the first SQL tuning set and two execution plans in the second SQL tuning set. Furthermore, the SQL statement was executed 22 times in the first SQL tuning set, but only 20 times in the second SQL tuning set.

The Plan Execution Statistics section shows runtime statistics per execution plan (or plan hash value). The Plans Before Change column lists plans and their associated execution statistics for the first SQL tuning set; the Plans After Change columns lists these values for the second SQL tuning set. Execution plans structures for both SQL tuning sets are shown at the end of the report.

You can use these sections in the report to identify changes in execution plans between two SQL tuning sets. This is important because changes in execution plans may be a result of test changes that can have a direct impact to performance. When comparing two SQL tuning sets, SQL Performance Analyzer reports execution plan changes when a SQL statement has:

• One plan in both SQL tuning sets, but the plan structure is different

• More than one plan, and the number of plans in both SQL tuning sets are:

  • The same, but at least one plan in the second SQL tuning set is different from all plans in the first SQL tuning set

  • Different

After evaluating the SQL statement and plan changes, determine if further action is required. If the SQL statement has regressed, perform one of the following actions:

• Tune the regressed SQL statement, as described in "Tuning Regressed SQL Statements Using APIs"

• Create SQL plan baselines, as described in "Creating SQL Plan Baselines Using APIs"

6.2.4 Tuning Regressed SQL Statements Using APIs

After reviewing the SQL Performance Analyzer report, you should tune any regressed SQL statements that are identified after comparing the SQL performance. If there are large numbers of SQL statements that appear to have regressed, you should try to identify the root cause and make system-level changes to rectify the problem. In cases when only a few SQL statements have regressed, consider using the SQL Tuning Advisor to implement a point solution for them, or creating SQL plan baselines to instruct the optimizer to select the original execution plan in the future.

To tune regressed SQL statements reported by SQL Performance Analyzer using APIs, create a SQL tuning task for the SQL Performance Analyzer execution by using the CREATE_TUNING_TASK function in the DBMS_SQLTUNE package:

BEGIN
  DBMS_SQLTUNE.CREATE_TUNING_TASK(
    spa_task_name => 'my_spa_task',
    spa_task_owner => 'immchan',
    spa_compare_exec => 'my_exec_compare');
  DBMS_SQLTUNE.EXECUTE_TUNING_TASK(spa_task_name => 'my_spa_task');
END;
/

This example creates and executes a SQL tuning task to tune the SQL statements that regressed in the compare performance execution named my_exec_compare of the SQL Performance Analyzer task named my_spa_task. In this case, it is important to use this version of the CREATE_TUNING_TASK function call. Otherwise, SQL statements may be tuned in the environment from the production system where they were captured, which will not reflect the system change.

Table 6-1 lists the SQL Performance Analyzer parameters that can be used with the DBMS_SQLTUNE.CREATE_TUNING_TASK function.

After tuning the regressed SQL statements, you should test these changes using SQL Performance Analyzer. Run a new SQL trial on the test system, followed by a second comparison (between this new SQL trial and the first SQL trial) to validate your results. Once SQL Performance Analyzer shows that performance has stabilized, implement the fixes from this step to your production system.

Starting with Oracle Database 11g Release 2, SQL Tuning Advisor performs an alternative plan analysis when tuning a SQL statement. SQL Tuning Advisor reviews the execution history of the SQL statement, including any historical plans stored in the Automatic Workload Repository. If SQL Tuning Advisor finds alternate plans, it allows you to choose a specific plan and create a plan baseline to ensure that the desired execution plan is used for that SQL statement.

6.2.5 Tuning Regressed SQL Statements From a Remote SQL Trial Using APIs

If you chose to execute the SQL workload remotely on a separate database, then you should tune any regressions identified by the SQL trials on the remote database, instead of the system where the SQL Performance Analyzer task resides.

To tune regressed SQL statements from a remote SQL trial:

1. On the system running SQL Performance Analyzer, create a subset of the regressed SQL statements as a SQL tuning set:

   DECLARE
     sqlset_cur  DBMS_SQLTUNE.SQLSET_CURSOR;
   BEGIN
     DBMS_SQLTUNE.CREATE_SQLSET('SUB_STS1', 'test purpose');
    
     OPEN sqlset_cur FOR
       SELECT value(p)
       FROM table(
         DBMS_SQLTUNE.SELECT_SQLPA_TASK(
           task_name  => 'SPA_TASK1',
           execution_name => 'COMP',
           level_filter => 'REGRESSED')) p;
      
     DBMS_SQLTUNE.LOAD_SQLSET('SUB_STS1', sqlset_cur);
    
     CLOSE sqlset_cur;
   END;
   /
   

   Other than 'REGRESSED', you can use other filters to select SQL statements for the SQL tuning set, such as 'CHANGED', 'ERRORS', or 'CHANGED_PLANS'. For more information, see Oracle Database PL/SQL Packages and Types Reference.

2. Create a staging table to where the SQL tuning set will be exported:

   BEGIN
     DBMS_SQLTUNE.CREATE_STGTAB_SQLSET(
       table_name  => 'STG_TAB1',
       schema_name => 'JOHNDOE',
       tablespace_name => 'TBS_1',
       db_version => DBMS_SQLTUNE.STS_STGTAB_11_1_VERSION);
   END;
   /
   

   Use the db_version parameter to specify the appropriate database version to where the SQL tuning set will be exported and tuned. In this example, the staging table will be created with a format so that it can be exported to a system running Oracle Database 11g Release 1, where it will later be tuned using SQL Tuning Advisor. For other database versions, see Oracle Database PL/SQL Packages and Types Reference for that release.

3. Export the SQL tuning set into the staging table:

   BEGIN
     DBMS_SQLTUNE.PACK_STGTAB_SQLSET(
       sqlset_name => 'SUB_STS1', 
       sqlset_owner => 'JOHNDOE', 
       staging_table_name => 'STG_TAB1', 
       staging_schema_owner => 'JOHNDOE', 
       db_version => DBMS_SQLTUNE.STS_STGTAB_11_1_VERSION);
   END;
   /
   

4. Move the staging table to the remote database (where the SQL workload was executed) using the mechanism of choice (such as Oracle Data Pump or database link).

5. On the remote database, import the SQL tuning set from the staging table:

   BEGIN
     DBMS_SQLTUNE.UNPACK_STGTAB_SQLSET(
       sqlset_name => 'SUB_STS1', 
       staging_table_name => 'STG_TAB1', 
       replace => TRUE);
   END;
   /
   

6. Tune the regressed SQL statements in the SQL tuning set by running SQL Tuning Advisor:

   BEGIN
     sts_name  := 'SUB_STS1';
     sts_owner := 'JOHNDOE';
     tune_task_name := 'TUNE_TASK1';
     tname := DBMS_SQLTUNE.CREATE_TUNING_TASK(sqlset_name  => sts_name, 
                                              sqlset_owner => sts_owner, 
                                              task_name    => tune_task_name);
     EXEC DBMS_SQLTUNE.SET_TUNING_TASK_PARAMETER(:tname, 
                                                 'APPLY_CAPTURED_COMPILENV', 
                                                 'FALSE');
     exec_name := DBMS_SQLTUNE.EXECUTE_TUNING_TASK(tname);
   END;
   /
   

   Note:

   The APPLY_CAPTURED_COMPILENV parameter used in this example is only supported by Oracle Database 11g Release 1 and newer releases. If you are testing a database upgrade from an earlier version of Oracle Database, SQL Tuning Advisor will use the environment variables stored in the SQL tuning set instead.

After tuning the regressed SQL statements, you should test these changes using SQL Performance Analyzer. Run a new SQL trial on the test system, followed by a second comparison (between this new SQL trial and the first SQL trial) to validate your results. Once SQL Performance Analyzer shows that performance has stabilized, implement the fixes from this step to your production system.

6.2.6 Creating SQL Plan Baselines Using APIs

Creating SQL plan baselines for regressed SQL statements with plan changes is another option to running the SQL Tuning Advisor. Doing so instructs the optimizer to use the original execution plans for these SQL statements in the future.

To create SQL plan baselines for the original plans, first create a subset of a SQL tuning set of only the regressed SQL statements. Next, create SQL plan baselines for this subset of SQL statements by loading their plans using the LOAD_PLANS_FROM_SQLSET function of the DBMS_SPM package, as shown in the following example:

DECLARE
  my_plans PLS_INTEGER;
BEGIN
  my_plans := DBMS_SPM.LOAD_PLANS_FROM_SQLSET(
    sqlset_name => 'regressed_sql');
END;
/

6.2.7 Using SQL Performance Analyzer Views

You can query the following views to monitor SQL Performance Analyzer and view its analysis results:

• The DBA_ADVISOR_TASKS and USER_ADVISOR_TASKS views display descriptive information about the SQL Performance Analyzer task that was created.

• The DBA_ADVISOR_EXECUTIONS and USER_ADVISOR_EXECUTIONS views display information about task executions. SQL Performance Analyzer creates at least three executions to analyze the SQL performance impact caused by a database change on a SQL workload. The first execution collects a pre-change version of the performance data. The second execution collects a post-change version of the performance data. The third execution performs the comparison analysis.

• The DBA_ADVISOR_FINDINGS and USER_ADVISOR_FINDINGS views display the SQL Performance Analyzer findings. SQL Performance Analyzer generates the following types of findings:

  • Problems, such as performance regression

  • Symptoms, such as when the structure of an execution plan has changed

  • Errors, such as nonexistence of an object or view

  • Informative messages, such as when the structure of an execution plan in the pre-change version is different than the one stored in the SQL tuning set

• The DBA_ADVISOR_SQLPLANS and USER_ADVISOR_SQLPLANS views display a list of all execution plans.

• The DBA_ADVISOR_SQLSTATS and USER_ADVISOR_SQLSTATS views display a list of all SQL compilations and execution statistics.

• The V$ADVISOR_PROGRESS view displays the operation progress of SQL Performance Analyzer. Use this view to monitor how many SQL statements have completed or are awaiting execution in a SQL trial. The SOFAR column indicates the number of SQL statements processed so far, and the TOTAL WORK column shows the total number of SQL statements to be processed by the task execution.

You must have the SELECT_CATALOG_ROLE role to access the DBA views.