Mohamed Houri’s Oracle Notes

January 27, 2018

Adaptive Cursor Sharing and parallel execution plan

Filed under: Oracle — hourim @ 7:33 am

I have shown and demonstrated at different places that there exist three pre-requisites for a cursor, using natural or forced bind variable, to be bind sensitive:

  • it uses a range (inequality) predicate
  • it uses an equality predicate and histogram
  • it uses a partition key in its predicate

I think that from now on we can add a fourth pre-requisite which is:

  • the very first execution of this cursor should not be run via a parallel execution plan.

Let’s demonstrate this point (the model should be taken from this post):

alter system flush shared_pool;
alter session set cursor_sharing= force;
alter table t_acs parallel;

select count(1) from t_acs where n2 = 1;

SQL_ID  7ck8k47bnqpnv, child number 0
-------------------------------------
------------------------------------------------
| Id  | Operation         | Name       | Rows  |
------------------------------------------------
|   0 | SELECT STATEMENT  |            |       |
|   1 |  SORT AGGREGATE   |            |     1 |
|*  2 |   INDEX RANGE SCAN| T_ACS_IDX1 |     1 |
------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - access("N2"=:SYS_B_1)

select
    child_number
   ,is_bind_sensitive
   ,is_bind_aware
 from 
   gv$sql
 where
    sql_id ='7ck8k47bnqpnv';

CHILD_NUMBER I I
------------ - -
           0 Y N

The very first execution of the above cursor used a serial execution plan and an equality predicate on a skewed column having Frequency histogram. This is why it has been marked bind sensitive.

But what would have happened to the cursor if its very first execution has been ran parallely?

alter system flush shared_pool;
select count(1) from t_acs where n2 = 1e6;
-----------------------------------------------------------
| Id  | Operation              | Name     | Rows  | Bytes |
-----------------------------------------------------------
|   0 | SELECT STATEMENT       |          |       |       |
|   1 |  SORT AGGREGATE        |          |     1 |     3 |
|   2 |   PX COORDINATOR       |          |       |       |
|   3 |    PX SEND QC (RANDOM) | :TQ10000 |     1 |     3 |
|   4 |     SORT AGGREGATE     |          |     1 |     3 |
|   5 |      PX BLOCK ITERATOR |          |  1099K|  3220K|
|*  6 |       TABLE ACCESS FULL| T_ACS    |  1099K|  3220K|
-----------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   6 - access(:Z>=:Z AND :Z<=:Z)
       filter("N2"=:SYS_B_1)

Note
-----
   - Degree of Parallelism is 8 because of table property

select
    child_number
   ,is_bind_sensitive
   ,is_bind_aware
 from 
   gv$sql
 where
    sql_id ='7ck8k47bnqpnv';

CHILD_NUMBER I I
------------ - -
           0 N N

Since the very first execution of the cursor uses a parallel execution plan Oracle refuses to set its bind sensitive property. And from this stage and on, until a cursor is flushed out from the library cache, all cursor executions will share the same parallel execution plan.

But what would have happened if the very first cursor execution would have used a serial plan?

alter system flush shared_pool;

select count(1) from t_acs where n2 = 100;
select count(1) from t_acs where n2 = 100;
select count(1) from t_acs where n2 = 100;
------------------------------------------------
| Id  | Operation         | Name       | Rows  |
------------------------------------------------
|   0 | SELECT STATEMENT  |            |       |
|   1 |  SORT AGGREGATE   |            |     1 |
|*  2 |   INDEX RANGE SCAN| T_ACS_IDX1 |   100 |
------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 – access("N2"=:SYS_B_1)

select
    child_number
   ,is_bind_sensitive
   ,is_bind_aware
   ,executions
 from 
   gv$sql
 where
    sql_id ='7ck8k47bnqpnv';

CHILD_NUMBER I I EXECUTIONS
------------ - - ----------
           0 Y N          3

As expected the cursor is bind sensitive. Let’s now make it bind aware :

select count(1) from t_acs where n2 = 1000;
select count(1) from t_acs where n2 = 1000;
select count(1) from t_acs where n2 = 1000;
select count(1) from t_acs where n2 = 1000;

select
    child_number
   ,is_bind_sensitive
   ,is_bind_aware
   ,executions
 from 
   gv$sql
 where
    sql_id ='7ck8k47bnqpnv';

CHILD_NUMBER I I EXECUTIONS
------------ - - ----------
           0 Y N          6
           1 Y Y          1

And here where the serious stuff starts. I will show you how Oracle will unset the bind sensitive and bind awareness property of the above cursor whenever the execution plan triggered by the ECS layer code and produced by CBO is a parallel plan:

select count(1) from t_acs where n2 = 1e6;

SQL_ID  7ck8k47bnqpnv, child number 2
-------------------------------------
-----------------------------------------------------------
| Id  | Operation              | Name     | Rows  | Bytes |
-----------------------------------------------------------
|   0 | SELECT STATEMENT       |          |       |       |
|   1 |  SORT AGGREGATE        |          |     1 |     3 |
|   2 |   PX COORDINATOR       |          |       |       |
|   3 |    PX SEND QC (RANDOM) | :TQ10000 |     1 |     3 |
|   4 |     SORT AGGREGATE     |          |     1 |     3 |
|   5 |      PX BLOCK ITERATOR |          |  1099K|  3219K|
|*  6 |       TABLE ACCESS FULL| T_ACS    |  1099K|  3219K|
-----------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   6 - access(:Z>=:Z AND :Z<=:Z)
       filter("N2"=:SYS_B_1)

Note
-----
   - Degree of Parallelism is 8 because of table property

select
    child_number
   ,is_bind_sensitive
   ,is_bind_aware
   ,executions
 from 
   gv$sql
 where
    sql_id ='7ck8k47bnqpnv';

CHILD_NUMBER I I EXECUTIONS
------------ - - ----------
           0 Y N          6
           1 Y Y          1
           2 N N          1

The new child number 2 is not anymore bind sensistive nor bind aware. That’s the most important conclusion of this article : Oracle doesn’t allow ACS to work with parallelism.

Summary

ACS has been implemented for very frequently used queries having different set of bind variables values each of which generating different amount of I/O. In this context, Oracle decided to cancel ACS whenever a parallel plan is triggred by ECS. Before 12cR2 there was a bug identified by Bug 16226575 in which ACS was disabled for query having decorated parallel object irrespective of the plan being chosen serial or parallel. As we saw in this article this has been fixed in 12cR2. ACS will be disabled only if its produced execution plan is parallel.

Advertisements

January 20, 2018

NESTED LOOP and full/right outer join in modern RDBMS

Filed under: CBO,explain plan — hourim @ 11:36 am

The Oracle Cost Based Optimizer, the MS-SQL Server optimizer and the PostgreSQL query planner cannot use a NESTED LOOP physical operation to execute FULL OUTER and RIGHT OUTER joins logical operations. They all address the RIGHT OUTER join limitation by switching the inner and the outer row source so that a LEFT OUTER JOIN can be used. While the first two optimizer turn the FULL OUTER join into a LEFT OUTER join concatenated with an ANTI-join, PostgreSQL query planner will always use a HASH/MERGE JOIN to do a FULL OUTER join.

Let’s make this less confusing by starting with the basics. The algorithm of a NESTED LOOP physical operation is:

for each row ro in the outer row source 
loop
   for each row ri in the inner row source
   loop
     if ro joins ri then return current row
   end loop
end loop

1.Oracle 12cR2

A simple execution of the above algorithm can be schematically represented via the following Oracle execution plan:

select /*+ use_nl(t1 t2) */ 
      t1.*
from t1 inner join t2
on t1.n1 = t2.n1;

----------------------------------------------------------------
| Id  | Operation          | Name   | Starts | E-Rows | A-Rows |
----------------------------------------------------------------
|   0 | SELECT STATEMENT   |        |      1 |        |      3 |
|   1 |  NESTED LOOPS      |        |      1 |      4 |      3 |
|   2 |   TABLE ACCESS FULL| T1     |      1 |      3 |      3 |
|*  3 |   INDEX RANGE SCAN | T2_IDX |      3 |      1 |      3 |
----------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   3 – access("T1"."N1"="T2"."N1")

As you can see, in accordance with the algorithm, for each row in T1 table (A-Rows=3 operation id n°2) we scanned 3 times (Starts = 3 operation id n°3) the T2_IDX index.

Let’s now try a FULL OUTER join but without any hint:

select  
      t1.*
from t1 full outer join t2
on t1.n1 = t2.n1;

---------------------------------------------------
| Id  | Operation             | Name     | Rows  |
---------------------------------------------------
|   0 | SELECT STATEMENT      |          |       |
|   1 |  VIEW                 | VW_FOJ_0 |     4 |
|*  2 |   HASH JOIN FULL OUTER|          |     4 |
|   3 |    TABLE ACCESS FULL  | T1       |     3 |
|   4 |    TABLE ACCESS FULL  | T2       |     4 |
---------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - access("T1"."N1"="T2"."N1")

So far so good. A HASH JOIN FULL OUTER to honor a full outer join between two tables.

But what if I want to use a NESTED LOOP FULL OUTER instead of HASH JOIN FULL OUTER join ?

select  /*+ use_nl(t1 t2) */
      t1.*
from t1 FULL outer join t2
on t1.n1 = t2.n1;

-------------------------------------------------------
| Id  | Operation            | Name   | Rows  | Bytes |
-------------------------------------------------------
|   0 | SELECT STATEMENT     |        |       |       |
|   1 |  VIEW                |        |     6 |   120 |
|   2 |   UNION-ALL          |        |       |       |
|   3 |    NESTED LOOPS OUTER|        |     4 |    40 |
|   4 |     TABLE ACCESS FULL| T1     |     3 |    21 |
|*  5 |     INDEX RANGE SCAN | T2_IDX |     1 |     3 |
|   6 |    NESTED LOOPS ANTI |        |     2 |    12 |
|   7 |     TABLE ACCESS FULL| T2     |     4 |    12 |
|*  8 |     TABLE ACCESS FULL| T1     |     2 |     6 |
-------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   5 - access("T1"."N1"="T2"."N1")
   8 - filter("T1"."N1"="T2"."N1")

What the heck is this execution plan of 8 operations?

Instead of having a simple NESTED LOOP FULL OUTER I got a concatenation of NESTED LOOPS OUTER and a NESTED LOOPS ANTI join.That’s an interesting transformation operated by the CBO.

Should I have tried to reverse engineer the query that sits behind the above execution plan I would have very probably obtained the following query:

select  
      t1.*
from 
      t1 
     ,t2 
where t1.n1 = t2.n1(+)
union all
select 
     t2.*
from t2
where not exists (select /*+ use_nl(t2 t1) */ 
                      null 
                  from t1 
                  where t1.n1 = t2.n1);

------------------------------------------------------
| Id  | Operation           | Name   | Rows  | Bytes |
------------------------------------------------------
|   0 | SELECT STATEMENT    |        |       |       |
|   1 |  UNION-ALL          |        |       |       |
|   2 |   NESTED LOOPS OUTER|        |     4 |    40 |
|   3 |    TABLE ACCESS FULL| T1     |     3 |    21 |
|*  4 |    INDEX RANGE SCAN | T2_IDX |     1 |     3 |
|   5 |   NESTED LOOPS ANTI |        |     2 |    20 |
|   6 |    TABLE ACCESS FULL| T2     |     4 |    28 |
|*  7 |    TABLE ACCESS FULL| T1     |     2 |     6 |
------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   4 - access("T1"."N1"="T2"."N1")
   7 - filter("T1"."N1"="T2"."N1")

In fact when I have directed Oracle to use a NESTED LOOP to FULL OUTER JOIN T1 and T2 it has turned out my instruction into:

 T1 LEFT OUTER JOIN T2 UNION ALL T2 ANTI JOIN T1

Which is nothing else than :

  • select all rows from T1 and T2 provided they join
  • add to these rows, rows from T1 that don’t join (LEFT OUTER)
  • add to these rows, all rows from T2 that don’t join (ANTI) with rows from T1

Do you know why Oracle did all this somehow complicated gymnastic?

It did it because I asked it to do an impossible operation: NESTED LOOP doesn’t support FULL OUTER join.

It doesn’t support RIGHT OUTER join as well as shown below:

select  /*+ use_nl(t1 t2) */
      t1.*
from t1 
RIGHT outer join t2
on t1.n1 = t2.n1;

---------------------------------------------------
| Id  | Operation          | Name | Rows  | Bytes |
---------------------------------------------------
|   0 | SELECT STATEMENT   |      |       |       |
|   1 |  NESTED LOOPS OUTER|      |     4 |    40 |
|   2 |   TABLE ACCESS FULL| T2   |     4 |    12 |
|*  3 |   TABLE ACCESS FULL| T1   |     1 |     7 |
---------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   3 - filter("T1"."N1"="T2"."N1")

Don’t be confused here. The above RIGHT OUTER JOIN has been turned into a LEFT OUTER JOIN by switching the inner and the outer table. As such, T2 being placed in the left side of the join Oracle is able to use a NESTED LOOP to operate a LEFT OUTER JOIN. You will see this clearly explained in the corresponding SQL Server execution plan I will show later in this article.

2. PostgreSQL 10.1

Since there are no hints in PostgreSQL to make a join using a NESTED LOOP I will start by cancelling hash and merge join operations as shown below:

postgres=# set enable_mergejoin=false;
SET
postgres=# set enable_hashjoin=false;
SET

And now I am ready to show you how the PostgreSQL query planner turns a right outer join into a left outer join when a NESTED LOOP operation is used:

postgres=# explain
postgres-# select
postgres-#       t1.*
postgres-# from t1 right outer join t2
postgres-# on t1.n1 = t2.n1;

                            QUERY PLAN
-------------------------------------------------------------------
 Nested Loop Left Join  (cost=0.00..95.14 rows=23 width=42)
   Join Filter: (t1.n1 = t2.n1)
   ->  Seq Scan on t2  (cost=0.00..1.04 rows=4 width=4)
   ->  Materialize  (cost=0.00..27.40 rows=1160 width=42)
         ->  Seq Scan on t1  (cost=0.00..21.60 rows=1160 width=42)
(5 lignes)

However, in contrast to Oracle and MS-SQL Server, PostgreSQL query planner is unable to transform a full outer join into a combination of an NESTED LOOP LEFT OUTER join and an ANTI-join as the following demonstrates:

explain 
select
      t1.*
from t1 full outer join t2
on t1.n1 = t2.n1;
                                QUERY PLAN
--------------------------------------------------------------------------
 Hash Full Join  (cost=10000000001.09..10000000027.27 rows=1160 width=42)
   Hash Cond: (t1.n1 = t2.n1)
   ->  Seq Scan on t1  (cost=0.00..21.60 rows=1160 width=42)
   ->  Hash  (cost=1.04..1.04 rows=4 width=4)
         ->  Seq Scan on t2  (cost=0.00..1.04 rows=4 width=4)

Spot in passing how disabling the hash join option (set enable_hashjoin=false) is not an irreversible action. Whenever the query planner is unable to find another way to accomplish its work it will use all option available even those being explicitely disabled.

3. MS-SQL Server 2016



4. Summary
In several if not all modern Relational DataBase Management Systems, NESTED LOOP operation doesn’t support right outer and full outer join. Oracle, MS-SQL Server and PostgreSQL turn “T1 right outer join T2” into “T2 left outer join T1” by switching the inner and the outer row source. Oracle and SQL Server turn a full outer join between T1 and T2 into a T1 left outer join T2 union-all T2 anti-join T1. PostgreSQL will always use a hash/merge to full outer join T1 and T2.

Model

--@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
-- NESTED LOOP and full/right outer join : Oracle
--@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
drop table t1;
drop table t2;
create table t1(t1_id int , t1_vc varchar2(10));

insert into t1 values (1, 't1x');
insert into t1 values (2, 't1y');
insert into t1 values (3, 't1z');
 
create index t1_idx on t1(t1_id);

create table t2 (t2_id int, t2_vc varchar(10));

insert into t2 values (2, 't2x');
insert into t2 values (3, 't2y');
insert into t2 values (3, 't2yy');
insert into t2 values (4, 't2z');

create index t2_idx on t2(t2_id);

December 28, 2017

Index statistics

Filed under: Statistics — hourim @ 8:06 pm

There are so many situations in the Oracle database world that I would have probably never encountered if there hadn’t been so many different real life applications with different problems that needed solving. A couple of days ago, a critical query of one of such applications, experienced severe performance deterioration. When I started looking at this query I’ve first decided to print out the statistics of its historical executions shown below:

SQL> @HistStats
Enter value for sql_id: gqj0qgtduh1k7
Enter value for from_date: 01122017

SNAP_BEGIN                PLAN_HASH_VALUE      EXECS  AVG_ETIME    AVG_PIO    AVG_LIO   AVG_ROWS
------------------------- --------------- ---------- ---------- ---------- ---------- ----------
01/12/17 12:00:44.897 AM        179157465        178          4        357     244223        543
01/12/17 01:00:50.192 AM        179157465        40           6       3976     246934        400
01/12/17 04:00:02.121 AM        179157465         2           2        171     246203        400
../..
03/12/17 05:00:22.064 AM        179157465        10           18      4231     245802        400
../..
04/12/17 09:00:19.716 AM        179157465        13           1        4       253433        482

The most important characteristic of this query is that it consumes 250K logical I/O per execution and that it can sometime last up to 18 seconds. This is a query generated by Hibernate. Unfortunately, due to the Hibernate obsession of changing table aliases, this query has got a new sql_id when it has been deployed in production via the December release. This is why we lost the track of the query pre-release statistics that would have allowed us to get the previous acceptable execution plan.

Starting from the end
Let’s, exceptionally, start this article by showing the statistics of this query after I have fixed it

SQL> @HistStats
Enter value for sql_id: gqj0qgtduh1k7
Enter value for from_date: 01122017

SNAP_BEGIN                PLAN_HASH_VALUE      EXECS  AVG_ETIME    AVG_PIO    AVG_LIO   AVG_ROWS
------------------------- --------------- ---------- ---------- ---------- ---------- ----------
01/12/17 12:00:44.897 AM        179157465        178          4        357     244223        543
01/12/17 01:00:50.192 AM        179157465        40           6       3976     246934        400
01/12/17 04:00:02.121 AM        179157465         2           2        171     246203        400
../..
03/12/17 05:00:22.064 AM        179157465        10           18      4231     245802        400
../..
04/12/17 09:00:19.716 AM        179157465        13           1        4       253433        482
../..
11/12/17 02:00:04.453 PM        1584349102        4           0        14       17825        482 –- fixed
12/12/17 12:00:38.622 AM        1584349102       67           1        661      13840        522 
12/12/17 01:00:42.740 AM        1584349102       40           0        103       7605        852 
13/12/17 05:00:27.270 AM        1584349102       10           1        845      16832        400 

Notice how, thanks to the fix I will explain in the next section, I have got a remarkable drop of the number of Logical I/O consumption from 244K to 17K per execution on average.

How this has been achieved?
The real time sql monitoring report of the degraded query confirms the bad response time of 12 seconds:

Global Information
------------------------------
 Status              :  DONE (ALL ROWS)                 
 Instance ID         :  2                    
 Session             :  xxxx (324:25737) 
 SQL ID              :  gqj0qgtduh1k7        
 SQL Execution ID    :  16777216             
 Execution Started   :  12/05/2017 06:01:20  
 First Refresh Time  :  12/05/2017 06:01:29  
 Last Refresh Time   :  12/05/2017 06:01:32  
 Duration            :  12s                  
 Module/Action       :  JDBC Thin Client/-           
 Service             :  xxxxx            
 Program             :  JDBC Thin Client/
 Fetch Calls         :  10

This report presents an oddity in the following part of its 54 execution plan operations:

---------------------------------------------------------------------------------------
| Id  | Operation                           | Name         | Starts | E-Rows | A-Rows |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |              |      1 |        |    398 |
|* 39 |  TABLE ACCESS BY INDEX ROWID BATCHED| T1           |      1 |     26 |    398 |
|* 40 |   INDEX RANGE SCAN                  | IDX_PRD_TYPE |      1 |     2M |     3M |
---------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
  39 - filter(("PRODUCT_ID"=:1 OR "PRODUCT_ID"=:2 OR "PRODUCT_ID"=:3 OR 
        "PRODUCT_ID"=:4 ../..  OR "PRODUCT_ID"=:398))
  40- access("PRODUCT_TYPE"='M'

This oddity resides in the fact that we went from 3M of rowids supplied by the index at operation 40 to only 398 rows at the table level. Such a situation can have one of the following explanations:

  • The CBO has made a very bad choice preferring the IDX_PRD_TYPE index over another existing index with a much better index effective selectivity.
  • IDX_PRD_TYPE is the sole index with which Oracle can cover the PRODUCT_ID and PRODUCT_TYPE predicates. And, in this case, it represents an imprecise index that should be re-designed if we want to efficiently cover the above query

The answer is straightforward If you remember that a column having an _ID suffix or prefix in its name refers generally to a primary key. The T1 table has indeed a primary key index defined as follows:

 T1_PK (PRODUCT_ID) 

In addition the 398 rows returned at the table level are nothing else than the 398 elements constituting the in-list of the corresponding query reproduced here below:

SELECT 
  *
FROM
   t1
WHERE 
   PRODUCT_ID 
       in (:1,:2,:3,:4,:5,....,:398)
AND PRODUCT_TYPE = 'M';

Table altered.

While the bad selected index has the following definition:

 IDX_PRD_TYPE(PRODUCT_TYPE,COLX)

And the initial question turned to : why the primary key index has not been used in this critical part of the executio plan?

At this stage of the investigations it becomes crystal clear for me that using the primary key index is, to a very large extent, better that using the index on the PRODUCT_TYPE column. This is simply because, if used by the CBO, the primary key index will supply its parent table operation by at most one rowid per element in the in-list predicate. That is for an in-list of 398 elements Oracle will only filter 398 rows at the table level using the PRODUCT_TYPE = ‘M’ filter. That’s largely better than filtering 3M of rows at the table level using the 3M of index rowids supplied by the IDX_PRD_TYPE(PRODUCT_TYPE,COLX) index.

Bear in mind as well that all parameters participating in the index desirability are at their default values. And that the PRODUCT_TYPE column has a remarkable frequency histogram showing perfectly the not evenly distribution of its 13 distinct values.

select 
   column_name
  ,histogram 
from 
  user_tab_col_statistics 
where 
  table_name = 'T1'
and
  column_name = 'PRODUCT_TYPE';

COLUMN_NAME  HISTOGRAM
------------ ---------------
PRODUCT_TYPE FREQUENCY

The explain plan for command applied when the primary key index is forced and when not used shows that the combined table/index access cost is cheaper when the IDX_PRD_TYPE index is used as the following proves:

-------------------------------------------------------------------
| Id  | Operation                    | Name  | Rows  | Cost (%CPU)|
-------------------------------------------------------------------
|   0 | SELECT STATEMENT             |       |    27 |   519   (1)|
|   1 |  INLIST ITERATOR             |       |       |            |
|*  2 |   TABLE ACCESS BY INDEX ROWID| T1    |    27 |   519   (1)|
|*  3 |    INDEX UNIQUE SCAN         | T1_PK |   398 |   399   (0)|
-------------------------------------------------------------------

--------------------------------------------------------------------------------
| Id  | Operation                           | Name         | Rows  |Cost (%CPU)|
--------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |              |    27 |  58   (0)|
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| T1           |    27 |  58   (0)|
|*  2 |   INDEX RANGE SCAN                  | IDX_PRD_TYPE |  2183K|   3   (0)|
--------------------------------------------------------------------------------

And the initial question turned to : why the cost of the inapropriate index is cheaper?

After few minutes of reflection I decided to get the statistics of these two indexes:

select
   index_name
  ,distinct_keys
  ,leaf_blocks
from
   user_indexes
where
  table_name = 'T1';
 
INDEX_NAME       DISTINCT_KEYS LEAF_BLOCKS
---------------- ------------- -----------
T1_PK                 32633730    111089
IDX_PRD_TYPE

And finally the initial question turned to be: why the IDX_PRD_TYPE index has no statistics?

The answer is:

  • The statistics of T1 table have been voluntarily locked since several months ago
  • The IDX_PRD_TYPE index has been very recently created and implemented in Production

The combination of the above two points resulted in a newly created index without any statistics making this index appearing very cheap in the CBO eyes and consequently leading to a very bad choice and to a noticeable performance issue.

Obviously gathering the statistics of this index makes it less desirable in favour of the primary key index as the following proves:

exec dbms_stats.unlock_table_stats(user, 't1');
begin
 dbms_stats.gather_index_stats(user, 'IDX_PRD_TYPE', no_invalidate =>false);
end;
/
exec dbms_stats.unlock_table_stats(user, 't1');

-------------------------------------------------------------------
| Id  | Operation                    | Name  | Rows  | Cost (%CPU)|
-------------------------------------------------------------------
|   0 | SELECT STATEMENT             |       |    27 |   519   (1)|
|   1 |  INLIST ITERATOR             |       |       |            |
|*  2 |   TABLE ACCESS BY INDEX ROWID| T1    |    27 |   519   (1)|
|*  3 |    INDEX UNIQUE SCAN         | T1_PK |   398 |   399   (0)|
-------------------------------------------------------------------

--------------------------------------------------------------------------------
| Id  | Operation                           | Name         | Rows  |Cost (%CPU)|
--------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |              |    27 |  137K   (2|
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| T1           |    27 |  137K   (2)|
|*  2 |   INDEX RANGE SCAN                  | IDX_PRD_TYPE |  2183K| 5678    (1)|
--------------------------------------------------------------------------------

Bottom Line

All things being equal when you decide to lock the statistics of one your tables then bear in mind that any newly created index on this table will not have any statistics. This index will the be, naturally, much more desirable than any other index created before the lock of the table statistics resulting, fairlly likely, into wrong index choice and bad execution plans. If you are in a similar situation you had then better to manually gather statistics of your new indexes as long as the statistics of your table are kept locked.

December 22, 2017

Null-Accepting Semi-Join

Filed under: CBO — hourim @ 11:15 am

Introduction

Null-Accepting semi-join is a new enhancement brought to the CBO by the 12cR1 release. It extends the semi-join algorithm to qualify rows from the table in the left hand side that have a null value in the join column.It kicks in for queries like the following one:

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
         OR exists (SELECT null FROM t2 where t2.n1 = t1.n1)
		);

It is recognisable in execution plans via its acronym NA (Null-Accepting)

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     6 |
|*  2 |   HASH JOIN SEMI NA |      |     7 |    42 |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|   4 |    TABLE ACCESS FULL| T2   |    10 |    30 |
----------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - access("T2"."N1"="T1"."N1")

It appears in the CBO 10053 trace file as (for its HASH JOIN version):

 Best:: JoinMethod: HashNullAcceptingSemi

Don’t get confused by the NA acronym that appears in the ANTI-JOIN operation where it refers to Null-Aware rather than to Null-Accepting as shown in the following execution plan and 10053 trace file respectively:

SELECT  
   count(1)
FROM t1
   WHERE t1.n1 NOT IN (select n1 from t2);

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     6 |
|*  2 |   HASH JOIN ANTI NA |      |     1 |     6 |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|   4 |    TABLE ACCESS FULL| T2   |    10 |    30 |
----------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 – access("T1"."N1"="N1")

 Best:: JoinMethod: HashNullAwareAnti

If you want to clear this confusion then remember that the Null-Accepting transformation occurs for rows that (SEMI) join while Null-Aware transformation is for rows that would not join (ANTI).

The semi-join Null-Accepting logical operation can also be serviced by the NESTED LOOP physical operation as the following demonstrates:

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
         OR exists (SELECT /*+ NL_SJ */ null FROM t2 where t2.n1 = t1.n1)
		); 

------------------------------------------------------
| Id  | Operation             | Name | Rows  | Bytes |
------------------------------------------------------
|   0 | SELECT STATEMENT      |      |       |       |
|   1 |  SORT AGGREGATE       |      |     1 |     6 |
|   2 |   NESTED LOOPS SEMI NA|      |     7 |    42 |
|   3 |    TABLE ACCESS FULL  | T1   |    10 |    30 |
|*  4 |    TABLE ACCESS FULL  | T2   |     7 |    21 |
------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   4 – filter("T2"."N1"="T1"."N1")

Best:: JoinMethod: NestedLoopNullAcceptingSemi

It is the ability, acquired by the CBO as from 12cR1, to unnest the above kind of disjunctive subquery that makes the Null-Accepting transformation possible as shown in the corresponding10053 trace file:

*****************************
Cost-Based Subquery Unnesting
*****************************
SU: Unnesting query blocks in query block SEL$1 (#1) that are valid to unnest.
Subquery Unnesting on query block SEL$1 (#1)
SU: Performing unnesting that does not require costing.
SU: Considering subquery unnest on query block SEL$1 (#1).
SU: Checking validity of unnesting subquery SEL$2 (#2)
SU: Transforming EXISTS subquery to a join.
Registered qb: SEL$5DA710D3 0x1d225e60 (SUBQUERY UNNEST SEL$1; SEL$2)

Prior to 12cR1 it was not possible to automatically unnest the above subquery to join it with its parent block leading to the below execution plan where the inner table T2 is scanned mutliple times:

SELECT /*+ gather_plan_statistics */ 
   count(1)
FROM t1
   WHERE(t1.n1 is null
         OR exists (SELECT /*+ no_unnest */ null FROM t2 where t2.n1 = t1.n1)
        );  
---------------------------------------------------------------
| Id  | Operation           | Name | Starts | E-Rows | A-Rows |
---------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |      1 |        |      1 |
|   1 |  SORT AGGREGATE     |      |      1 |      1 |      1 |
|*  2 |   FILTER            |      |      1 |        |     10 |
|   3 |    TABLE ACCESS FULL| T1   |      1 |     10 |     10 |
|*  4 |    TABLE ACCESS FULL| T2   |      7 |      1 |      7 |
---------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter(("T1"."N1" IS NULL OR  IS NOT NULL))
   4 - filter("T2"."N1"=:B1)

The Bug

The Null-Accepting semi-join transformation comes, unfortunately, with a bug already identified in MyOracle Support via number 21201446. Here’s below how it can be reproduced:


SQL> select banner from v$version where rownum = 1;

BANNER
--------------------------------------------------------------------------------
Oracle Database 12c Enterprise Edition Release 12.1.0.2.0 - 64bit Production

create table t1 as select rownum n1, trunc((rownum -1/5)) n2 from dual connect by level <= 10; 

create table t2 as select rownum n1, trunc((rownum -1/3)) n2 from dual connect by level <= 10;

update t1 set n1 = null where n1 in (5,6,7);

exec dbms_stats.gather_table_stats(user, 't1');

exec dbms_stats.gather_table_stats(user, 't2');

SQL> SELECT  
         count(1)
    FROM t1
      WHERE(t1.n1 is null
         OR exists 
           (SELECT null FROM t2 where nvl(t2.n1,42) = nvl(t1.n1,42))
	    );

  COUNT(1)
----------
         7

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     6 |
|*  2 |   HASH JOIN SEMI NA |      |     7 |    42 |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|   4 |    TABLE ACCESS FULL| T2   |    10 |    30 |
----------------------------------------------------
 
Predicate Information (identified by operation id):
---------------------------------------------------
 2 - filter(NVL("T2"."N1",42)=NVL("T1"."N1",42))

Using the HASH JOIN physical operation the query returns 7 rows. But it returns 10 rows when it uses the NESTED LOOP operation a shown below:

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
          OR exists 
           (SELECT /*+ NL_SJ */ null FROM t2 where nvl(t2.n1,42) = nvl(t1.n1,42))
	   );

  COUNT(1)
----------
        10
		
------------------------------------------------------
| Id  | Operation             | Name | Rows  | Bytes |
------------------------------------------------------
|   0 | SELECT STATEMENT      |      |       |       |
|   1 |  SORT AGGREGATE       |      |     1 |     6 |
|   2 |   NESTED LOOPS SEMI NA|      |     7 |    42 |
|   3 |    TABLE ACCESS FULL  | T1   |    10 |    30 |
|*  4 |    TABLE ACCESS FULL  | T2   |     7 |    21 |
------------------------------------------------------
 
Predicate Information (identified by operation id):
---------------------------------------------------
   4 – filter(NVL("T2"."N1",42)=NVL("T1"."N1",42))	

The Null-Accepting semi-join transformation is driven by the following hidden parameter which , if cancelled, will workarround this bug as shown below:

SQL> alter session set "_optimizer_null_accepting_semijoin"=false;

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
          OR exists 
           (SELECT null FROM t2 where nvl(t2.n1,42) = nvl(t1.n1,42))
	   );

  COUNT(1)
----------
        10

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     3 |
|*  2 |   FILTER            |      |       |       |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|*  4 |    TABLE ACCESS FULL| T2   |     1 |     3 |
----------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter(("T1"."N1" IS NULL OR  IS NOT NULL))
   4 - filter(NVL("T2"."N1",42)=NVL(:B1,42))

How this has been fixed in 12cR2?

Because of the bug n° 21201446 It seems that Oracle has completely cancelled the Null-Accepting semi-join transformation in 12cR2 for both NESTED LOOP and HASH JOIN physical operations when the NVL function is applied on the join column. Here’s below why I am thinking so:

SQL> select banner from v$version where rownum=1;

BANNER
-----------------------------------------------------------------------------
Oracle Database 12c Enterprise Edition Release 12.2.0.1.0 - 64bit Production

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
          OR exists 
           (SELECT null FROM t2 where nvl(t2.n1,42) = nvl(t1.n1,42))
	   );

  COUNT(1)
----------
        10

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     3 |
|*  2 |   FILTER            |      |       |       |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|*  4 |    TABLE ACCESS FULL| T2   |     1 |     3 |
----------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter(("T1"."N1" IS NULL OR  IS NOT NULL))
   4 – filter(NVL("T2"."N1",42)=NVL(:B1,42))

In 12cR1, as we’ve shown above, the NESTED LOOP was not concerned by the bug. As from 12cR2 the NESTED LOOP SEMI JOIN is not anymore allowed to occur if a NVL function is applied on the join column as the following proves:

SELECT  
   count(1)
FROM t1
   WHERE(t1.n1 is null
          OR exists 
           (SELECT /*+ NL_SJ */ null FROM t2 where nvl(t2.n1,42) = nvl(t1.n1,42))
	   );

  COUNT(1)
----------
        10

----------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes |
----------------------------------------------------
|   0 | SELECT STATEMENT    |      |       |       |
|   1 |  SORT AGGREGATE     |      |     1 |     3 |
|*  2 |   FILTER            |      |       |       |
|   3 |    TABLE ACCESS FULL| T1   |    10 |    30 |
|*  4 |    TABLE ACCESS FULL| T2   |     1 |     3 |
----------------------------------------------------

Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1
   3 - SEL$1 / T1@SEL$1
   4 - SEL$2 / T2@SEL$2

Outline Data
-------------
  /*+
      BEGIN_OUTLINE_DATA
      IGNORE_OPTIM_EMBEDDED_HINTS
      OPTIMIZER_FEATURES_ENABLE('12.1.0.1')
      DB_VERSION('12.2.0.1')
      OPT_PARAM('_optimizer_null_accepting_semijoin' 'false')
      ALL_ROWS
      OUTLINE_LEAF(@"SEL$2")
      OUTLINE_LEAF(@"SEL$1")
      FULL(@"SEL$1" "T1"@"SEL$1")
      PQ_FILTER(@"SEL$1" SERIAL)
      FULL(@"SEL$2" "T2"@"SEL$2")
      END_OUTLINE_DATA
  */

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter(("T1"."N1" IS NULL OR  IS NOT NULL))
   4 - filter(NVL("T2"."N1",42)=NVL(:B1,42))

December 10, 2017

Compress enable novalidate

Filed under: Oracle — hourim @ 7:47 pm

We knew here that we can’t DDL optimise a BASIC compressed table. So I uncompressed my table in a 12.2.0.1.0 and DDL optimise it as in the following:

SQL> select 
        pct_free
	,compression
	,compress_for
    from 
	 user_tables
    where 
	   table_name = 'T1';

  PCT_FREE COMPRESS COMPRESS_FOR
---------- -------- -------------
         0 DISABLED

SQL> alter table t1 add c_ddl number default 42 not null;

Table altered.

But I changed my mind and finally decided to revert back and drop this newly added column:

SQL> alter table t1 drop column c_ddl;
alter table t1 drop column c_ddl
                           *
ERROR at line 1:
ORA-39726: unsupported add/drop column operation on compressed tables

Damned database!!!

What’s going on here? My table is not compressed. Is it?

In fact I didn’t show you how I have uncompressed the t1 table:

SQL> alter table t1 nocompress;

Table altered.

Compressing or un-compressing a table without moving it will do nothing to its data. When the COMPRESSION column of a table is DISABLED this doesn’t necessarily mean that this table doesn’t contain compressed data . And this is probably the reason why the above bizarre error pops up out from nowhere.
In fact in order to have a significant meaning, the COMPRESSION column should normally always be considered with the PCT_FREE column (notice in passing that this is why I printed above the pct_free value):

SQL> select
         table_name
        ,compression
        ,pct_free
        ,compress_for
     from 
     user_tables
    where table_name = 'T1';

TABLE_NAME COMPRESS   PCT_FREE COMPRESS_FOR
---------- -------- ---------- --------------
T1         DISABLED          0

When PCT_FREE equals zero this is generally a hint that Oracle is considering BASIC compression since it thinks that the table is not considered for updates anymore. The definition given by Oracle to the COMPRESSION column goes in the same direction as well:

COMPRESSION	 	Indicates whether table compression 
                    is enabled (ENABLED) or not (DISABLED)

This is crystal clear: COMPRESSION referes to the status of the current compression. It doesn’t tell anything about the compression history of the table if any.

Compressing/uncompressing a table without moving it is nothing else than implementing or desimplementing compression but only for newly inserted rows. This is why I think that Oracle could have avoided the confusion by using the same command as when dealing with table constraints. Instead of this:

SQL> alter table t1 nocompress;

We would have this :

SQL> alter table t1 nocompress enable novalidate;

And of course the COMPRESSION column should have had a status indicating whether it is validated or not.

In the 10046 file of a compressed table with a move of its data and without a move we can see this respectively:

alter table t1 move compress basic

call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.01       0.02          0          0          1           0
Execute      1      1.43       1.44          3       1865       2548     1000000
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      1.45       1.46          3       1865       2549     1000000

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106  
Number of plan statistics captured: 1

Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         0          0          0  LOAD AS SELECT  T1 (cr=1905 pr=1 pw=1382 time=1436345 us
   1000000    1000000    1000000   TABLE ACCESS FULL T1 (cr=1797 pr=0 pw=0 time=104378 

Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  PGA memory operation                         1140        0.00          0.01
  Disk file operations I/O                        1        0.00          0.00
  db file sequential read                         3        0.00          0.00
  direct path write                               1        0.00          0.00
  reliable message                                4        0.00          0.00
  enq: RO - fast object reuse                     2        0.00          0.00
  enq: CR - block range reuse ckpt                2        0.00          0.00
  SQL*Net message to client                       1        0.00          0.00
  SQL*Net message from client                     1        5.93          5.93
********************************************************************************
alter table t1 nocompress


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          2           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.00       0.01          0          0          2           0

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106  

Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  Compression analysis                            8        0.00          0.00
  SQL*Net message to client                       1        0.00          0.00
  SQL*Net message from client                     1        6.65          6.65
********************************************************************************

Spot that whileOracle managed to rework a million rows of in the first case it didn’t touch any in the second case proving that nocompress/compress without a move doesn’t concern existing data.

Since it concerns only new data then an inserted row should be compressed/non-compressed. Let’s see:

SQL> create table t1 as select
      rownum n1
     ,mod(rownum,10) n2
     from dual
     connect by level <=1e6;

Table created.

SQL> @sizeBysegNameMB
Enter value for segment_name: t1
Enter value for owner: c##mhouri


SEGMENT_TYPE  TABLESPACE_NAME  SEGMENT_NAME PARTITION_NAME          MB
------------- ---------------- ------------ --------------- ----------
TABLE         USERS            T1                                   15
                                                            ----------
Total Segment Size                                                  15

Let’s compress this table without moving its data and get its new size:

SQL> alter table t1 compress;

Table altered.

SQL> @sizeBysegNameMB
Enter value for segment_name: t1
Enter value for owner: c##mhouri


SEGMENT_TYPE  TABLESPACE_NAME  SEGMENT_NAME PARTITION_NAME          MB
------------- ---------------- ------------ --------------- ----------
TABLE         USERS            T1                                   15
                                                            ----------
Total Segment Size                                                  15

As you can see the size of the table remains intact indicating again that compressing without moving does nothing to the existing data.

Let’s now direct path load a single row, dump the corresponding block and analyse the dump to see if compression has been activated or not:

SQL> select rowid, n1,n2 from t1 where n1=1;

ROWID                      N1         N2
------------------ ---------- ----------
AAATYNAAHAAAdpTAAA          1          1

SQL> insert /*+ append */ into t1 select 1,42 from dual;

1 row created.

SQL> select rowid, n1,n2 from t1 where n1=1;
select rowid, n1,n2 from t1 where n1=1
                         *
ERROR at line 1:
ORA-12838: cannot read/modify an object after modifying it in parallel


SQL> commit;

Commit complete.

SQL> select rowid, n1,n2 from t1 where n1=1;

ROWID                      N1         N2
------------------ ---------- ----------
AAATYNAAHAAAdpTAAA          1          1
AAATYNAAHAADOgcAAA          1         42

SQL> select
  2    dbms_rowid.rowid_relative_fno('AAATYNAAHAADOgcAAA') file_no
  3  , dbms_rowid.rowid_block_number('AAATYNAAHAADOgcAAA') block_no
  4  from dual;

   FILE_NO   BLOCK_NO
---------- ----------
         7     845852


SQL> alter session set tracefile_identifier='DumpingDirectLoadRowInsert';

Session altered.

SQL> alter system dump datafile 7 block 845852;

System altered.

Unfortunately I couldn’t spot anything interesting in the dump file which indicates that the inserted row has been compressed. I will probably come back to this article in a near future.

December 7, 2017

Optimizer statistics gathering

Filed under: direct path — hourim @ 8:24 pm

As from Oralce 12cR1, the Optimizer can automatically collect statistics on empty tables provided they are, notably, direct path loaded. This is recognisable in execution plans through the new row source operation named OPTIMIZER STATISTICS GATHERING

-----------------------------------------------------------
| Id  | Operation                        | Name   | Rows  |
-----------------------------------------------------------
|   0 | INSERT STATEMENT                 |        |  1000K|
|   1 |  LOAD AS SELECT                  | T1_TAR |       |
|   2 |   OPTIMIZER STATISTICS GATHERING |        |  1000K|
|   3 |    TABLE ACCESS FULL             | T1     |  1000K|
-----------------------------------------------------------

This is, in principle, a good initiative. But look at what I have been asked to trouble shoot a couple of days ago:

Global Information
------------------------------
 Status              :  EXECUTING                 
 Instance ID         :  1                    
 Session             :  xxxx (15:32901) 
 SQL ID              :  am6923m45s203        
 SQL Execution ID    :  16777216             
 Execution Started   :  12/02/2017 08:01:17  
 First Refresh Time  :  12/02/2017 08:01:28  
 Last Refresh Time   :  12/05/2017 11:55:51  
 Duration            :  272276s                  
 Module/Action       :  SQL*Plus/-           
 Service             :  xxxxx            
 Program             :  sqlplus@xxxx

===================================================================================================
| Id |             Operation               |  Name  | Activity | Activity Detail                  |
|    |                                     |        |   (%)    |   (# samples)                    |
===================================================================================================
| -> 0 | INSERT STATEMENT                  |        |   0.00   | Data file init write             |   
| -> 1 |   LOAD AS SELECT                  | T1_TAR |   0.06   | Cpu(28)                          |
|                                          |        |          | direct path write(2)             |
| -> 2 |    OPTIMIZER STATISTICS GATHERING |        |  99.93   | Cpu(52956)                       |
| -> 3 |     TABLE ACCESS FULL             | T1     |          |  SQL*Net more data from dblink(1)|               
==================================================================================================    

You have to believe me when I say that this insert had been running for 3 days when the above SQL report was taken. As you can see 99% of the direct path load execution time was spent gathering statistics at operation in line n°2. This is how a new feature designed to help starts disturbing you.

The insert/select statement selects a CLOB from a remote database. I was curious to see whether getting rid of this CLOB from the insert operation would make things better:

SQL>set timing on
 
    INSERT
   /*+ append */
 INTO T1_TAR
    (col1
   -- ,clob_col2
	,col3
	,col4
	,col5
	,col6
	,col7
	,col8
	)
 SELECT
    t1.col1 
   --,t1.clob_col2
   ,t1.col3
   ,t1.col4
   ,t1.col5
   ,t1.col6
   ,t1.col7
   ,t1.col8
  FROM
    t1@dblink t1;

338481182 rows created.

Elapsed: 00:11:38.85

------------------------------------------------------------
| Id  | Operation          | Name   | Rows  | Bytes |IN-OUT|
-----------------------------------------------------------
|   0 | INSERT STATEMENT   |        |  318M|    28GB|      |   
|   1 |  LOAD AS SELECT    | T1_TAR |      |        |      |
|   2 |   REMOTE           | T1     |  318M|    28GB|R->S  |
------------------------------------------------------------

That’s really interesting.

Why would the presence or the absence of a CLOB column in the insert/select statement allow or prevent the CBO from gathering statistics online?

In fact the CLOB column has nothing to do with the online statistics gathering decision. Should I have commented any other column I would have had the same behaviour as the following prooves:

INSERT
   /*+ append */
 INTO T1_TAR
    (col1
    ,clob_col2
	,col3
	,col4
	,col5
	,col6
	--,col7
	,col8
	)
 SELECT
    t1.col1 
   ,t1.clob_col2
   ,t1.col3
   ,t1.col4
   ,t1.col5
   ,t1.col6
   --,t1.col7
   ,t1.col8
  FROM
    t1@dblink t1;
------------------------------------------------------------
| Id  | Operation          | Name   | Rows  | Bytes |IN-OUT|
-----------------------------------------------------------
|   0 | INSERT STATEMENT   |        |  318M|    28GB|      |   
|   1 |  LOAD AS SELECT    | T1_TAR |      |        |      |
|   2 |   REMOTE           | T1     |  318M|    28GB|R->S  |
------------------------------------------------------------

The OPTIMIZER STATISTICS GATHERING operation requires the presence of the totality of the table columns in order to kick in during direct path load operations. This is probably because Oracle doesn’t want to collect statistics for a bunch of columns and keep others without statistics. This being said, this is another non-really documented restriction that comes with this new 12c feature.

Since I was not going to challenge my client getting rid of one column from the insert/select I finally opted for locally cancelling the online statistics gathering using the corresponding hint as shown below:

INSERT
   /*+ append 
       no_gather_optimizer_statistics
   */
 INTO T1_TAR
    (col1
    ,clob_col2
    ,col3
    ,col4
    ,col5
    ,col6
    ,col7
    ,col8
    )
 SELECT
    t1.col1 
   ,t1.clob_col2
   ,t1.col3
   ,t1.col4
   ,t1.col5
   ,t1.col6
   ,t1.col7
   ,t1.col8
  FROM
    t1@dblink t1;

Global Information
------------------------------
 Status              :  DONE                 
 Instance ID         :  1                    
 Session             :  xxxx (132:63271) 
 SQL ID              :  Od221zf0srs6q        
 SQL Execution ID    :  16777216             
 Execution Started   :  12/02/2017 15:10:16  
 First Refresh Time  :  12/02/2017 15:10:22  
 Last Refresh Time   :  12/05/2017 15:32:56  
 Duration            :  1360s                  
 Module/Action       :  SQL*Plus/-           
 Service             :  xxxxx            
 Program             :  sqlplus@xxxx

SQL Plan Monitoring Details (Plan Hash Value=2098243032)
================================================================
| Id |             Operation   |  Name  |   Time    |   Rows   |
|    |                         |        | Active(s) | (Actual) |
================================================================
|  0 | INSERT STATEMENT        |        |      1355 |        2 |
|  1 |   LOAD AS SELECT        | T1_TAR |      1355 |        2 |
|  2 |     REMOTE              | T1     |      1355 |     338M |
================================================================    

That’s how we fixed this issue.

Watch out your direct path load over empty tables selecting from either local or distant databases; the new 12c online table and columns statics gathering might considerably delay your treatement.

November 23, 2017

Compressed tables and DDL optimisation

Filed under: compression — hourim @ 8:15 pm

DDL optimisation is a very useful feature with which adding a column having a default value to a very very big table is almost instantaneous. And, as an icing on the cake, this feature comes free of license.

Free of license have I said?

Yes. Until you hit the followings in a 12.2.0.1.0 release

SQL> alter table t1 add c_ddl number default 42 not null;
alter table t1 add c_ddl number default 42 not null
                   *
ERROR at line 1:
ORA-39726: unsupported add/drop column operation on compressed tables

My real life very very big table was BASIC compressed as is the t1 table


SQL> select
       table_name
      ,compression
      ,pct_free
      ,compress_for
    from 
      user_tables
    where 
      table_name = 'T1';

TABLE_NAME COMPRESS   PCT_FREE COMPRESS_FOR
---------- -------- ---------- -------------
T1         ENABLED           0 BASIC
 

If you want to DDL optimise a BASIC compressed table you have to switch for the paid advanced compression as the following demonstrates:


SQL> alter table t1 move compress for oltp;

Table altered.

SQL> select
       table_name
      ,compression
      ,pct_free
      ,compress_for
    from 
      user_tables
    where 
      table_name = 'T1';

TABLE_NAME COMPRESS   PCT_FREE COMPRESS_FOR
---------- -------- ---------- -------------
T1         ENABLED           0 ADVANCED

SQL> alter table t1 add c_ddl number default 42 not null;

Table altered.

SQL> select count(1) from t1 where c_ddl=42;

  COUNT(1)
----------
   1000000

SQL_ID  59ysfcmfx0s7t, child number 0
-------------------------------------
---------------------------------------------------
| Id  | Operation          | Name | Rows  | Bytes |
---------------------------------------------------
|   0 | SELECT STATEMENT   |      |       |       |
|   1 |  SORT AGGREGATE    |      |     1 |     8 |
|*  2 |   TABLE ACCESS FULL| T1   | 10000 | 80000 |
---------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter(NVL("C_DDL",42)=42)

 

With a compressed table using advanced compression I succeeded to add a column to table t1 having a default value using the DDL optimisation feature. The presence of the NVL function in the predicate part of the above execution plan is what allows me to say that, indeed, the DDL optimisation feature has been used during the alter table command.

This is how Oracle implicitely force you to buy a new license. If you want to DDL optimise a compressed table than you have to use the OLTP paid option.

However, If you are in the same situation as that of my customer and you don’t want to pay for extra license fees then proceed as shown below:


SQL> alter table t1 move nocompress parallel 8;

Table altered.

SQL> alter table t1 add c_ddl number default 42 not null;

Table altered.

SQL> alter table t1 move compress basic parallel 8;

Table altered.

SQL> alter index idx_t1_pk rebuild parallel 8;

Index altered.

SQL> alter index idx_t1_pk noparallel;

Index altered.

Summary

In contrast to the free BASIC compression Advanced compression allows DDL optimisation. Take this into account when deciding about the type of compression you will choose at design time.

Model

create table t1 as select 
rownum n1
,mod(rownum,10) n2
from dual
connect by level <=1e6;

create unique index idx_t1_pk on t1(n1);

alter table t1 add constraint t1_pk primary key(n1) using index;

October 28, 2017

Cursor selectivity cube -Part II

Filed under: Oracle — hourim @ 1:40 pm

In the same vein as the preceding blog post, in this second and last post pf the series I will provide three differents scripts to a bind aware cursor that owes its bind sensitiveness property from a bind variable value having a Hybrid histogram. The first one gives the selectivity cube of a popular Hybrid histogram value. The second script do the same thing for a non-popular Hybrid histogram having an endpoint number. The third and last script gives the selectivity cube of a non captured Hybrid histogram value.

1. Cursor Selectivity cube for a popular Hybrid histogram

In order to put all this in action I am going to use the model I have found in this article:

SQL> desc acs_test_tab
 Name                Null?    Type
 ------------------- -------- -------------
 ID                  NOT NULL NUMBER
 RECORD_TYPE                  NUMBER
 DESCRIPTION                  VARCHAR2(50)

SQL> alter session set cursor_sharing=force;

SQL> select
       column_name
      ,num_distinct
      ,num_buckets
      ,sample_size
      ,histogram
    from
       user_tab_col_statistics
    where table_name = 'ACS_TEST_TAB'
    and column_name  = 'RECORD_TYPE';

COLUMN_NAME  NUM_DISTINCT NUM_BUCKETS SAMPLE_SIZE HISTOGRAM
------------ ------------ ----------- ----------- ------------
RECORD_TYPE         50080         254        5407 HYBRID

As you can see the RECORD_TYPE column has a HYBRID histogram with the following popular and non-popular values distribution:

select
         endpoint_number
        ,endpoint_actual_value
        ,endpoint_repeat_count
        --,bucket_size
        ,case when Popularity > 0 then 'Pop'
                   else 'Non-Pop'
          end Popularity
    from
   (
     select
         uth.endpoint_number
        ,uth.endpoint_actual_value
        ,uth.endpoint_repeat_count
        ,ucs.sample_size/ucs.num_buckets bucket_size      
        ,(uth.endpoint_repeat_count - ucs.sample_size/ucs.num_buckets) Popularity
    from
        user_tab_histograms uth
       ,user_tab_col_statistics ucs
   where
        uth.table_name   = ucs.table_name
    and uth.column_name   = ucs.column_name
    and uth.table_name    = 'ACS_TEST_TAB'
    and uth.column_name   = 'RECORD_TYPE'
    )
   order by endpoint_number;

ENDPOINT_NUMBER ENDPOINT_A ENDPOINT_REPEAT_COUNT POPULAR
--------------- ---------- --------------------- -------
              1 1                              1 Non-Pop
           2684 2                           2683 Pop     -- we will use this
           2695 569                            1 Non-Pop -- we wiil use this
           2706 1061                           1 Non-Pop
           2717 1681                           1 Non-Pop
           2727 1927                           1 Non-Pop
../..
           5364 98501                          1 Non-Pop
           5375 98859                          1 Non-Pop
           5386 99187                          1 Non-Pop
           5396 99641                          1 Non-Pop
           5407 99999                          1 Non-Pop

254 rows selected.

There are 254 endpoint actual values of which only one value is popular(2). The following query will be used all over this article to show how we can compute the cursor selectivity cube for the three types of Hybrid histogram mentioned in the introduction:

-- as the ambassador of popular values
SELECT MAX(id) FROM acs_test_tab WHERE record_type = 2;

We will start by getting the selectivity cube of the popular value 2 as shown below (script can be found at the end of this blog post):

SQL> @CurSelCubeHybridPop

PL/SQL procedure successfully completed.

Enter value for bind: 2

BIND              LOW       HIGH
---------- ---------- ----------
2             ,446588    ,545829

Now that we have figured out what will be the value of the cursor selectivity cube of a bind aware cursor using the popular value 2, let’s see whether we have been right or wrong:

SELECT MAX(id) FROM acs_test_tab WHERE record_type = 1;-- only for ACS warmup

SELECT MAX(id) FROM acs_test_tab WHERE record_type = 2;
SELECT MAX(id) FROM acs_test_tab WHERE record_type = 2;                            

SQL_ID  9vu42gjuudpvj, child number 1
-------------------------------------
---------------------------------------------------
| Id  | Operation          | Name         | Rows  |
---------------------------------------------------
|   0 | SELECT STATEMENT   |              |       |
|   1 |  SORT AGGREGATE    |              |     1 |
|*  2 |   TABLE ACCESS FULL| ACS_TEST_TAB | 49621 |
---------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 – filter("RECORD_TYPE"=:SYS_B_0)

select
    child_number
    ,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='9vu42gjuudpvj';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           1          0 0.446588   0.545829

Spot how the low-high value of child cursor n°1 matches perfectly the low and high value given by the CurSelCubeHybridPop.sql script.

2. Cursor Selectivity cube for a non-popular Hybrid histogram value having an endpoint number

Let’s now consider a non-popular value having an endpoint number in the histogram table and let’s figure out what would be the selectivity cube of its underlying bind aware cursor:

SQL> @CurSelCubeHybridNonPop

PL/SQL procedure successfully completed.

PL/SQL procedure successfully completed.

Enter value for bind: 569

BIND              LOW       HIGH
---------- ---------- ----------
569           ,000166    ,000203

And now we are ready to execute the corresponding query, get its execution plan and present the low and high value of the bind aware cursor when ran against this Hybrid non-popular value:

SELECT MAX(id) FROM acs_test_tab WHERE record_type = 569;

SQL_ID  9vu42gjuudpvj, child number 2 –- new execution plan
-------------------------------------
-----------------------------------------------------------------------------------
| Id  | Operation                            | Name                       | Rows  |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                     |                            |       |
|   1 |  SORT AGGREGATE                      |                            |     1 |
|   2 |   TABLE ACCESS BY INDEX ROWID BATCHED| ACS_TEST_TAB               |    18 |
|*  3 |    INDEX RANGE SCAN                  | ACS_TEST_TAB_RECORD_TYPE_I |    18 |
-----------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   3 – access("RECORD_TYPE"=:SYS_B_0)

select
    child_number
    ,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='9vu42gjuudpvj';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           2          0 0.000166   0.000203  –- new execution plan
           1          0 0.446588   0.545829

Spot again the precision of the forecast. The low and high value of child cursor n°2 correspond exactly to the selectivity cube of the Hybrid non-popular value anticipated by the CurSelCubeHybridNonPop.sql script.

3. Cursor Selectivity cube for a non-popular Hybrid histogram value without an endpoint number

A Hybrid histogram value without an endpoint number is a value that exists for the column but which has not been captured by the Histogram gathering program for reasons I am not going to expose here. We can get all those values via an appropriate query. 41 is one value among the not captured ones. Let’s use it in the following demonstration:

Firt we will get its expected selectivity cube:

SQL> @CurSelCubeHybridNonPopWithoutEndPoint

PL/SQL procedure successfully completed.

PL/SQL procedure successfully completed.

Enter value for bind: 41

        41        LOW       HIGH
---------- ---------- ----------
        41    ,000009    ,000011

This selectivity range is not included in the selectivity range of child cursor n°1 nor in that of child cursor n°2. This is why if we use it in the following query it will certainly force ECS to hard parse a new execution plan as the following proves:

SQL> SELECT MAX(id) FROM acs_test_tab WHERE record_type = 41;

SQL_ID  9vu42gjuudpvj, child number 3 -- new execution plan
-------------------------------------
-----------------------------------------------------------------------------------
| Id  | Operation                            | Name                       | Rows  |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                     |                            |       |
|   1 |  SORT AGGREGATE                      |                            |     1 |
|   2 |   TABLE ACCESS BY INDEX ROWID BATCHED| ACS_TEST_TAB               |     1 |
|*  3 |    INDEX RANGE SCAN                  | ACS_TEST_TAB_RECORD_TYPE_I |     1 |
-----------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   3 - access("RECORD_TYPE"=:SYS_B_0)

As it has been again correctly expected a new execution plan (child cursor n°3) has been compiled. But since the new execution plan is identical to an existing one (child cursor n°2) Oracle has merged the selectivities of these two cursors, kept shareable the last compiled one and deleted the old plan as the following proves:

select
    child_number
    ,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='9vu42gjuudpvj';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           3          0 0.000009   0.000203   -- new plan with merge selectivity
           2          0 0.000166   0.000203
           1          0 0.446588   0.545829

select
   child_number
  ,is_shareable
from
   gv$sql
where
   sql_id = '9vu42gjuudpvj';

CHILD_NUMBER I
------------ -
           0 N
           1 Y
           2 N –- deleted 
           3 Y – selectivity cube merged

4. Conclusion

The first blog post of this series provided a script with which we can anticipate the cursor selectivity cube range of a bind aware cursor when its bind sensitivenes is due to a predicate having a Frequency histogram. In this article we presented three new scripts giving the same anticipation for a bind aware cursor that owes its bind sensitivenes respectively to a popular Hybrid histogram, a non-popular Hybrid histogram having and endpoint number and a non captured Hybrid histogram. This concludes the series
CurSelCubeHybridNonPop

CurSelCubeHybridNonPopWithoutEndPoint

CurSelCubeHybridPop

October 25, 2017

Cursor selectivity cube -Part I

Filed under: Oracle — hourim @ 6:00 pm

Bind variable selectivity is the building block on which the Extended Cursor Sharing Layer code reasons to compile a new good enough execution plan or share an existing one. It kicks in only for a bind aware cursor. The underlying child cursor is given a selectivity interval comprised between a low and a high value derived from the bind variable selectivity that initiates it. This is what Oracle refers to as a cursor selectivity cube shown in the following picture:

The ECS layer code launches the bind-aware cursor matching algorithm at each soft parse of a bind aware cursor. If the new bind variable value selectivity is outside an existing selectivity cube (low-high exposed in gv$sql_cs_selectivity) then a new hard parse is done and a new child cursor with a new selectivity cube is created. If, however, the peeked bind variable selectivity falls into a range of an existing child cursor selectivity cube, ECS will then share the corresponding child cursor’s execution plan. Finally if a new hard parsed execution plan is equivalent to an existing one then both child cursors will be merged. The selectivity cube of the last created child cursor will be adjusted while the previous cursor which served the merge process will be marked un-shareable in order to save space in the memory and reduce the time spent during cursor pruning activity.

The rest of this article shows, first, how the selectivity cube (low-high value) is computed for a bind variable value with a Frequency histogram. It then explains how two cursors with the same execution plan but different selectivity cubes are merged to form a single child cursor with an updated low-high range interval.

1. Cursor Selectivity cube

For simplicity’s sake I am going to use my old and helpful model:

create table t_acs(n1 number, n2 number);

begin
for j in 1..1200150 loop
   if j=1 then
      insert into t_acs values (j,1);
   elsif j> 1 and j <= 101 then          
    insert into t_acs values (j,100);     
    elsif j> 101 and j <= 1101 then          
    insert into t_acs values (j,1000);     
    elsif j> 10001 and j <= 110001 then          
    insert into t_acs values (j,10000);     
    else insert into t_acs values (j,1000000);     
   end if;  end loop;  commit;  
end  
/  
create index t_acs_idx1 on t_acs(n2); 
begin 
dbms_stats.gather_table_stats
        (user, 't_acs' 
        ,method_opt => 'for all columns size skewonly'
        ,cascade => true
	    ,estimate_percent => dbms_stats.auto_sample_size);
end;
/

The above data set contains 1,200,150 rows of which the n2 column has 5 distinct highly skewed values as shown below:

SQL> select n2, count(1) from t_acs group by n2 order by 2 ;

        N2   COUNT(1)
---------- ----------
         1          1
       100        100
      1000        910
     10000     100000
   1000000    1099139

The n2 column has a FREQUENCY histogram.

SQL> select column_name, histogram
     from user_tab_col_statistics
     where table_name ='T_ACS'
     and column_name  ='N2';

 COLUMN_NAM HISTOGRAM
---------- -----------
N2         FREQUENCY

The selectivity of the numeric n2 column is then computed via the following formula:

SQL> select
        endpoint_actual_value bind
       ,round( endpoint_actual_value/1200150,6) bind_sel
       ,endpoint_number
       ,endpoint_number –(lag(endpoint_number,1,0)
                         over(order by endpoint_number)) value_count
     from user_tab_histograms
     where table_name ='T_ACS'
     and column_name  ='N2'
     order by endpoint_number;

BIND         BIND_SEL ENDPOINT_NUMBER VALUE_COUNT
---------- ---------- --------------- -----------
1             ,000001               1           1
100           ,000083             101         100
1000          ,000833            1011         910
10000         ,008332          101011      100000
1000000       ,833229         1200150     1099139

The cursor selectivity cube is computed using the selectivity of the n2 bind variable value and an offset of +- 10% far from that selectivity forming  the x and y abscises of the cursor selectivity cube(see the above figure):

SQL> select
    bind
   ,round((sel_of_bind - offset),6) low
   ,round((sel_of_bind + offset),6) high
from
   (select
      bind
     ,value_count/1200150 sel_of_bind
	 ,0.1*(value_count/1200150) offset
     from
     (
      select
        endpoint_actual_value bind
       ,round( endpoint_actual_value/1200150,6) bind_sel
       ,endpoint_number
       ,endpoint_number –(lag(endpoint_number,1,0)
                         over(order by endpoint_number)) value_count
     from user_tab_histograms
     where table_name ='T_ACS'
     and column_name  ='N2'
     order by endpoint_number
     )
    )
    ;

BIND              LOW       HIGH
---------- ---------- ----------
1             ,000001    ,000001
100           ,000075    ,000092
1000          ,000682    ,000834
10000         ,074991    ,091655
1000000       ,824251   1,007418

Let’s put this select into a sql script and name it CurSelCubeFreq.sql.

We will be back to this script later in this article. For this moment, we will put it on hold and we will embark on the cursor merge section.

2. Cursor merge

The cursor of the query I am going to execute is not yet bind aware as the following proves:

select
    child_number
	,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='42wmc4buuh6wb';

no rows selected

But the next execution will mark it bind aware (cursor sharing parameter is set to FORCE) and will generate a new (full table scan) execution plan:

SQL> select count(1) from t_acs where N2 = 1e6;

  COUNT(1)
----------
   1099139

SQL> start xpsimp

SQL_ID  42wmc4buuh6wb, child number 1
-------------------------------------
--------------------------------------------
| Id  | Operation          | Name  | Rows  |
--------------------------------------------
|   0 | SELECT STATEMENT   |       |       |
|   1 |  SORT AGGREGATE    |       |     1 |
|*  2 |   TABLE ACCESS FULL| T_ACS |  1099K|
--------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - filter("N2"=:SYS_B_1)

select
    child_number
	,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='42wmc4buuh6wb';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           1          0 0.824251   1.007418

And, for the sake of clarity, let’s print again the content of the CurSelCubeFre.sql (slightly updated):

SQL> @CurSelCubeFreq
Enter value for bind: 1e6

BIND              LOW       HIGH
---------- ---------- ----------
1000000       ,824251   1,007418

Spot the coincidence 🙂

Suppose now that I want to know whether a less selective bind variable value (1) will force a new hard parse or share an existing execution plan. For that, I will first get the selectivity cube of this bind variable as shown below:

-- The selectivity cube of bind variable 1
SQL> @CurSelCubeFreq
Enter value for bind: 1

BIND              LOW       HIGH
---------- ---------- ----------
1             ,000001    ,000001

As you can see this bind variable value has a selectivity outside outside of that of the existing child cursor n°1. This is why ECS will fairly likely trigger a new hard parse as the following proves:

SQL> select count(1) from t_acs where N2 = 1;

  COUNT(1)
----------
         1

SQL> start xpsimp

SQL_ID  42wmc4buuh6wb, child number 2 -- new execution plan
-------------------------------------
------------------------------------------------
| Id  | Operation         | Name       | Rows  |
------------------------------------------------
|   0 | SELECT STATEMENT  |            |       |
|   1 |  SORT AGGREGATE   |            |     1 |
|*  2 |   INDEX RANGE SCAN| T_ACS_IDX1 |     1 |
------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - access("N2"=:SYS_B_1)

select
    child_number
	,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='42wmc4buuh6wb';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           2          0 0.000001   0.000001  --> new execution plan
           1          0 0.824251   1.007418

Notice that we have got, as expected, a new child cursor n°2 with a new range of selectivity that has produced an index range scan execution plan.

Finally, what if, in the next run, I will use a bind variable value(10000) having a different selectivity but producing the same index range scan execution plan?

SQL> @CurSelCubeFreq
Enter value for bind: 10000

BIND              LOW       HIGH
---------- ---------- ----------
10000         ,074991    ,091655

SQL> select count(1) from t_acs where N2 = 10000;

  COUNT(1)
----------
    100000

SQL> start xpsimp

SQL_ID  42wmc4buuh6wb, child number 3
-------------------------------------
------------------------------------------------
| Id  | Operation         | Name       | Rows  |
------------------------------------------------
|   0 | SELECT STATEMENT  |            |       |
|   1 |  SORT AGGREGATE   |            |     1 |
|*  2 |   INDEX RANGE SCAN| T_ACS_IDX1 |   100K|
------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 - access("N2"=:SYS_B_1)

select
    child_number
	,range_id
   ,low
   ,high
 from
   gv$sql_cs_selectivity
 where
    sql_id ='42wmc4buuh6wb';

CHILD_NUMBER   RANGE_ID LOW        HIGH
------------ ---------- ---------- ----------
           3          0 0.000001   0.091655  --> new child cursor
           2          0 0.000001   0.000001
           1          0 0.824251   1.007418

Notice how the low value of child cursor n°3 (0.000001) corresponds to the low value of child cursor n°2 and not to the low selectivity of the bind variable value for which it has been compiled (.074991). This is because the selectivity of child cursor n°2 has been merged with that of child cursor n°3 since their execution plans are identical. While the selectivity cube of child cursor n°3 has been enlarged child cursor n°2 has been deleted (put in a non-shareable status) as shown below:

select
   child_number
  ,is_bind_aware
  ,is_shareable
 from
   gv$sql
 where
   sql_id ='42wmc4buuh6wb'; 

CHILD_NUMBER I I
------------ - -
           0 N N
           1 Y Y
           2 Y N → is not shareable
           3 Y Y → includes the selectivity of child n°2

If we want to know whether the bind variable value 100 will share an existing execution plan or force a new one we have to check if its selectivity falls into an existing child cursor selectivity cube or not:

SQL> @CurSelCubeFreq
Enter value for bind: 100

BIND              LOW       HIGH
---------- ---------- ----------
100           ,000075    ,000092

This is going to share the child cursor n°3 since its selectivity falls into the low-high range of that cursor:

SQL> select count(1) from t_acs where N2 = 100;

  COUNT(1)
----------
       100

SQL> start xpsimp

SQL_ID  42wmc4buuh6wb, child number 3
-------------------------------------
------------------------------------------------
| Id  | Operation         | Name       | Rows  |
------------------------------------------------
|   0 | SELECT STATEMENT  |            |       |
|   1 |  SORT AGGREGATE   |            |     1 |
|*  2 |   INDEX RANGE SCAN| T_ACS_IDX1 |   100K|
------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
   2 – access("N2"=:SYS_B_1)

3. Conclusion

That is how the Extended Cursor Sharing layer code works. A combination of bind variable selectivities, with a possibly extendable cursor selectivity cube, allows Oracle to decide, at each execution of a bind aware cursor, whether to share an existing execution plan, compile a brand new one, or merge two cursors to form a single one to the detriment of a deleted one. This last action reduces the memory usage and the number of child cursors during the non-innocent child cursor pruning that occurs when a shareable parent cursor with multiple childs is soft parsed.

October 20, 2017

DDL(constraint), DML(predicate) and SQL CASE

Filed under: SQL — hourim @ 7:36 pm

Check and integrity constraints are DDL (Data Definition Language) operations. They consider TRUE and NULL alike. Query predicates are DML (Data Manipulation Language) operations. They consider FALSE and NULL alike.

Did you spot the difference?

In this article Jonathan Lewis turns this to:

  • DDL predicate allows a row if it does not evalute to FALSE – which means it is allowed to evaluate to TRUE or NULL.
  • DML predicate returns a row if it evaluates to TRUE – which means it will not return a row if it evaluates to FALSE or NULL.

In this article I am going to show that using the SQL CASE statement we can make DDL and DML predicates behave identically.

The standard boolean logic difference between DDL and DML predicates having been already explained (see this and this excellent book) I am not going to repeat that demonstration here. But, for simplicity’s sake, I am going to reuse a model from one of those articles as shown below:

create table t1 (v1 varchar2(1));
alter table t1 add constraint t1_ck_v1 check (v1 in ('a','b','c'));
insert into t1 values ('a');
insert into t1 values ('b');
insert into t1 values ('c');

The t1_ck_v1 DDL check constraint is supposed to prevent insertions into t1 table of any value that doesn’t match ‘a’ or ‘b’ or ‘c’. But since DDL check and integrity constraint considers TRUE and NULL alike, we are allowed to insert a null value in t1 as the following demonstrates:

SQL> insert into t1 values (null);
1 row created.

It suffices to change the check constraint t1_ck_v1 definition and nulls will definitively be banned from entering into t1 table as shown below:

truncate table t1;
alter table t1 drop constraint t1_ck_v1;

alter table t1 add constraint t1_ck_v1 check
    (case when v1 in ('a','b','c') then 1 else 0 end = 1);

insert into t1 values ('a');
insert into t1 values ('b');
insert into t1 values ('c');

SQL> insert into t1 values (null);
insert into t1 values (null)
*
ERROR at line 1:
ORA-02290: check constraint (C##MHOURI.T1_CK_V1) violated

In contrast to the first created check constraint the new refactored one has successfully pre-empted a null value from being inserted into t1 table (we will see shortly why).

The above demonstrates that using the CASE statement we made a DDL predicates considering NULL and FALSE alike while they are originally implemented to treat TRUE and NULL alike. Let’s now check how the CASE statement works with a DML predicate:

SQL> select count(1) from t1;

  COUNT(1)
----------
         3

SQL> select count(1) from t1
     where
        ( case when v1 in ('a','b','c') then 1 else 0 end = 1);

  COUNT(1)
----------
         3

Indeed the case statement works the same way in both DML and DDL predicates.

And now the legitimate question is why does the CASE statement make the check constraint, which is a DDL predicate, consider FALSE and NULL alike?
The answer can be found in the Oracle official documentation:

“In a searched CASE expression, Oracle searches from left to right until it finds an occurrence of condition that is true, and then returns return_expr. If no condition is found to be true, and an ELSE clause exists, Oracle returns else_expr. Otherwise, Oracle returns null”

Thanks to the ELSE clause of the CASE-WHEN combination we transformed a NULL possibility into a NOT NULL value (0 in the current situation). This is why we should always forsee an ELSE into a CASE statement. Should we have ommited the ELSE clause in the last t1_ck_v1 we would have then allowed null values to be inserted into t1 table as shown below:

truncate table t1;
alter table t1 drop constraint t1_ck_v1;

alter table t1 add constraint t1_ck_v1 check
    (case when v1 in ('a','b','c') then 1 end = 1);

SQL> insert into t1 values(null);
1 row created.

Furthermore ommitting the ELSE in the CASE allows not null values not in the triplet (‘a’,’b’,’c’) to be inserted into t1 table as shown below:

SQL> insert into t1 values ('z');
1 row created.

SUMMARY
Protect yourself against the always threatening NULL by setting your column not null whenever possible. Bear in mind that the CASE statement, when used with its ELSE part, can not only make DDL predicate treats FALSE and NULL alike, but it behaves the same way during DML and DDL predicate.

Next Page »

Create a free website or blog at WordPress.com.

Tony's Oracle Tips

Tony Hasler's light hearted approach to learning about Oracle

Richard Foote's Oracle Blog

Focusing Specifically On Oracle Indexes, Database Administration and Some Great Music

Hatem Mahmoud Oracle's blog

Just another Oracle blog : Database topics and techniques

Mohamed Houri’s Oracle Notes

Qui se conçoit bien s’énonce clairement

Oracle Diagnostician

Performance troubleshooting as exact science

Raheel's Blog

Things I have learnt as Oracle DBA

Coskan's Approach to Oracle

What I learned about Oracle

So Many Oracle Manuals, So Little Time

“Books to the ceiling, Books to the sky, My pile of books is a mile high. How I love them! How I need them! I'll have a long beard by the time I read them”—Lobel, Arnold. Whiskers and Rhymes. William Morrow & Co, 1988.

EU Careers info

Your career in the European Union

Carlos Sierra's Tools and Tips

Tools and Tips for Oracle Performance and SQL Tuning

Oracle Scratchpad

Just another Oracle weblog

OraStory

Dominic Brooks on Oracle Performance, Tuning, Data Quality & Sensible Design ... (Now with added Sets Appeal)