Monday Feb 11, 2013

How do adaptive cursor sharing and SQL Plan Management interact?

We've received a lot of questions about how adaptive cursor sharing (ACS) and SQL plan management (SPM) interact.  We discussed this briefly in one of the original SPM posts, but in this post, we'll explain the concepts of how the two features interact, and show an example. 

The simplest way to reason about the interaction is to remember that they are responsible for two different tasks.  ACS controls whether or not a child cursor is shared on a particular execution.  For each execution of the query, ACS considers the current bind values and decides if an existing child cursor can be shared or if the optimizer should be given the chance to find a better plan for the current bind values.  SPM controls which plans the optimizer may choose.  If a child cursor is bind-aware, the decision to share or not is made irrespective of whether the query is controlled by SPM.  But once the query and its current bind values are sent to the optimizer for optimization, SPM constrains the optimizer's choice of plans, without regard to whether this query is being optimized due to ACS.

[Read More]

Wednesday Jan 09, 2013

How does SQL Plan Management match SQL statements to SQL plan baselines?

Happy New Year to all of our readers!

As more and more of your systems migrate to Oracle Database 11g, SQL Plan Management (SPM) is becoming increasingly popular as a mechanism to maintain plan stability and subsequently system performance. However, one of the key frustrations folks encounter when using SPM is that the SQL plan baseline they painstakingly created is not used.

[Read More]

Tuesday May 08, 2012

What is the difference between SQL Profiles and SQL Plan Baselines?

Since Oracle Database 11g was released I have gotten a lot of questions about the difference between SQL profiles and SQL plan baselines and why SQL profiles can be shared but SQL plan baselines can't. So I thought it would be a good idea to write a post explaining the differences between them and how they interact. But first let's briefly recap each feature.

The query optimizer normally uses information like object and system statistics, compilation environment, bind values and so on to determine the best plan for a SQL statement. In some cases, defects in either these inputs or the optimizer can lead to a sub-optimal plan.  A SQL profile contains auxiliary information that mitigates this problem.  When used together with its regular inputs, a SQL profile helps the optimizer minimize mistakes and thus more likely to select the best plan.

A SQL plan baseline for a SQL statement consists of a set of accepted plans. When the statement is parsed, the optimizer will only select the best plan from among this set. If a different plan is found using the normal cost-based selection process, the optimizer will add it to the plan history but this plan will not be used until it is verified to perform better than the existing accepted plan and is evolved. We described this behavior in more detail in a previous post.

[Read More]

Monday Oct 24, 2011

Does the use of SQL Plan Management and the DBMS_SPM database package require a tuning or diagnostic pack license?

Recently during the Open World conference a lot of people asked me about what additional licenses are need to use SQL Plan Management. No additional licenses are needed to use SQL Plan Management (SPM) or any of the procedures in the DBMS_SPM package. SPM is available as part of Oracle Database Enterprise Edition 11g.


I believe the confusion arises from the wording in the Oracle Database Licensing documentation  that says the Tuning Pack includes "Automatic Plan Evolution of SQL Plan Management". What this is actually referring to is the interaction between the SQL Tune Advisor's nightly tuning task and SPM. This interaction does require you to have a license to use the Tuning pack.

In Oracle Database 11g, the SQL Tuning Advisor, runs automatically during the maintenance window. This automatic SQL tuning task targets high-load SQL statements. These statements are identified by the execution performance data collected in the Automatic Workload Repository (AWR) snapshots and not from SPM. If the SQL Tuning Advisor finds a better execution plan for a SQL statement it will recommend a SQL profile. Some of these high-load SQL statements may already have SQL plan baselines created for them. If a SQL profile recommendation is implemented, the execution plan found by the SQL Tuning Task will be added as an accepted plan to the existing SQL plan baseline. 


The Tuning pack or RAT pack are also required if you want to load plans from a SQL Tuning Set (STS) into SPM. Again this is because an STS can only be created if you have licensed one of these packs. It has nothing to do with SPM.

Maria Colgan+

Tuesday Jul 12, 2011

How do I migrate stored outlines to SQL Plan Management?

Stored outlines were deprecated in Oracle Database 11g and even though we have done a number of posts about SQL Plan Management (SPM) we have never discussed the reasons why you should migrate stored outlines to SQL plan baselines and the steps necessary to do the migration. It is strongly recommended that you migrate existing stored outlines to SPM because:

[Read More]

Tuesday Feb 15, 2011

How do I deal with a third party application that has embedded hints that result in a sub-optimal execution plan in my environment?

I have gotten many variations on this question recently as folks begin to upgrade to Oracle Database 11g and there have been several posts on this blog and on others describing how to use SQL Plan Management (SPM) so that a non-hinted SQL statement can use a plan generated with hints. But what if the hint is supplied in the third party application and is causing performance regressions on your system?

[Read More]

Tuesday Nov 17, 2009

What should I do with old hints in my workload?

We promised in an earlier post to cover some of the questions from the Optimizer round table discussion at Oracle Open World. Here's our first in a series of posts that will address these questions.

Q: When moving from 10g to 11g, should hints in existing SQL be removed?

A: I was glad to see this question at the round table, since I think this is something people often overlook when adding hints to an application. Over time, hints can become stale. You may have added a hint to a query because of a weakness in an old version of the optimizer, but the hint might no longer be necessary to get the best plan. Even worse, the hint may force a plan which is suboptimal, for a couple of reasons:
  • Depending on how your data change over time, the plan choice that the hint enforces might no longer be a good choice. This problem can occur even without a database upgrade. For example, if the distribution of values in a column change over time, an old access path hint may no longer be appropriate.
  • The hint might prevent new optimizations from taking place. So after an upgrade, you will be stuck using the old hint-enforced plan, when the optimizer could have chosen something better. For example, you may have hinted a plan for a query with bind variables, to avoid a bad plan choice due to bind peeking in the pre-11g optimizer. In 11g, you will not get the benefits of adaptive cursor sharing because of the hints in the query.
These are just some of the reasons why it is good to re-evaluate the hints in your application from time to time. You can test this out using the _optimizer_ignore_hints parameter. Setting this parameter to TRUE will cause the optimizer to ignore the hints embedded in queries. You can set this on the session level, run your workload, and compare the performance to your baseline performance (with the hints). I recommend using SQL Performance Analyzer (SPA) to do this. Read the SPA white paper for more information on how to do that.

If you find that some of your queries really do still require hints to get an optimal plan, you should consider creating a SQL plan baseline, and allowing SPM to manage the plan for you. If you do this, then you will get the hinted plan for now, but if a better plan comes along later (for one of the reasons mentioned earlier), you will eventually get the benefits of the new plan, using the SPM evolve process. There are two ways that you can create a SQL plan baseline based on your hints that will be applied to the query without hints. The first way requires that a SQL plan baseline already exist for the unhinted query. That's described in an earlier post. Here are the steps to use if you don't already have a SQL plan baseline for the unhinted query.

1. Run the query with hints, and confirm that the plan is what you want:
var pid number
exec :pid := 100;
select /*+ leading(t) */ p.prod_name, s.amount_sold, t.calendar_year
from sales s, products p, times t
where s.prod_id = p.prod_id
and s.time_id = t.time_id
and p.prod_id < :pid; 
PROD_NAME                                          AMOUNT_SOLD CALENDAR_YEAR
-------------------------------------------------- ----------- -------------
...
select sql_id from v$sql where sql_text like 'select /*+ leading(t) */ p.prod_name%';
SQL_ID
-------------
2qtu6hy4rf1j9
select * from table(dbms_xplan.display_cursor(sql_id=>'2qtu6hy4rf1j9', 
                                              format=>'basic note'));
EXPLAINED SQL STATEMENT:
------------------------
select /*+ leading(t) */ p.prod_name, s.amount_sold, t.calendar_year from sales s, products p, times t where s.prod_id = p.prod_id and s.time_id = t.time_id and p.prod_id < :pid

Plan hash value: 2290436051
---------------------------------------------------------------
| Id  | Operation                            | Name           |
---------------------------------------------------------------
|   0 | SELECT STATEMENT                     |                |
|   1 |  HASH JOIN                           |                |
|   2 |   HASH JOIN                          |                |
|   3 |    TABLE ACCESS FULL                 | TIMES          |
|   4 |    PARTITION RANGE ALL               |                |
|   5 |     TABLE ACCESS BY LOCAL INDEX ROWID| SALES          |
|   6 |      BITMAP CONVERSION TO ROWIDS     |                |
|   7 |       BITMAP INDEX RANGE SCAN        | SALES_PROD_BIX |
|   8 |   TABLE ACCESS BY INDEX ROWID        | PRODUCTS       |
|   9 |    INDEX RANGE SCAN                  | PRODUCTS_PK    |
---------------------------------------------------------------

2. Load SQL plan baseline from cursor cache, and associate it with the unhinted query:

var sqltext clob;

begin
:sqltext := 'select p.prod_name, s.amount_sold, t.calendar_year
from sales s, products p, times t
where s.prod_id = p.prod_id
and s.time_id = t.time_id
and p.prod_id > :pid';
end;
/

exec :pls := dbms_spm.load_plans_from_cursor_cache( -
              sql_id => '2qtu6hy4rf1j9', -
              plan_hash_value => 2290436051, -
              sql_text => :sqltext);

3. Run the query without hints, and check that the SQL plan baseline was used.

select p.prod_name, s.amount_sold, t.calendar_year
from sales s, products p, times t
where s.prod_id = p.prod_id
and s.time_id = t.time_id
and p.prod_id < :pid;

PROD_NAME                                          AMOUNT_SOLD CALENDAR_YEAR
-------------------------------------------------- ----------- -------------
...

select * from table(dbms_xplan.display_cursor(sql_id=>'a1ax3265pq8x7',
                                              format=>'basic note'));

EXPLAINED SQL STATEMENT:
------------------------
select p.prod_name, s.amount_sold, t.calendar_year from sales s,
products p, times t where s.prod_id = p.prod_id and s.time_id =
t.time_id and p.prod_id < :pid

Plan hash value: 2290436051



---------------------------------------------------------------
| Id  | Operation                            | Name           |
---------------------------------------------------------------
|   0 | SELECT STATEMENT                     |                |
|   1 |  HASH JOIN                           |                |
|   2 |   HASH JOIN                          |                |
|   3 |    TABLE ACCESS FULL                 | TIMES          |
|   4 |    PARTITION RANGE ALL               |                |
|   5 |     TABLE ACCESS BY LOCAL INDEX ROWID| SALES          |
|   6 |      BITMAP CONVERSION TO ROWIDS     |                |
|   7 |       BITMAP INDEX RANGE SCAN        | SALES_PROD_BIX |
|   8 |   TABLE ACCESS BY INDEX ROWID        | PRODUCTS       |
|   9 |    INDEX RANGE SCAN                  | PRODUCTS_PK    |
---------------------------------------------------------------
Note ----- - SQL plan baseline SQL_PLAN_4rw2dhryc2w5h888547d3 used for this statement

Monday Feb 02, 2009

SQL Plan Management (Part 4 of 4): User Interfaces and Other Features

In the first three parts of this article, we have seen how SQL plan baselines are created, used and evolved. In this final installment, we will show some user interfaces, describe the interaction of SPM with other features and answer some of your questions.[Read More]

Monday Jan 26, 2009

SQL Plan Management (Part 3 of 4): Evolving SQL Plan Baselines

In the example in Part 2, we saw that the optimizer used an accepted plan instead of a brand new plan. The statement has two plans in its plan history, but only one is accepted and thus in the SQL plan baseline:[Read More]

Tuesday Jan 20, 2009

SQL Plan Management (Part 2 of 4) SPM Aware Optimizer

In Part 1, we saw how you can create SQL plan baselines. After you create a SQL plan baseline for a statement, subsequent executions on that statement will use the SQL plan baseline. From all the plans in the SQL plan baseline, the optimizer will select the one with the best cost in the current environment (including bind values, current statistics, parameters, etc.). The optimizer will also generate the best-cost plan that it would otherwise have used without a SQL plan baseline. However, this best-cost plan will not be used but instead added to the statement's plan history for later verification. In other words, the optimizer will use a known plan from the SQL plan baseline instead of a new and hitherto unknown plan. This guarantees no performance regression.[Read More]

Thursday Jan 08, 2009

SQL Plan Management (Part 1 of 4) Creating SQL plan baselines

Do you ever experience performance regressions because an execution plan has changed for the worse? If you have, then we have an elegant solution for you in 11g called SQL Plan Management (SPM). The next four posts on our blog will cover SPM in detail. Let's begin by reviewing the primary causes for plan changes.[Read More]
About

The Oracle Optimizer blog is written by members of the Optimizer development team. The goal of this blog is to provide an insight into the workings of the Optimizer and the statistics it relies on. The views expressed on this blog are our own and do not necessarily reflect the views of Oracle and its affiliates. The views and opinions expressed by visitors on this blog are theirs solely and may not reflect ours.

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today