[20190823]关于CPU成本计算2.txt

[20190823]关于CPU成本计算2.txt

--//前几天探究CPU cost时遇到的问题,获取行成本时我的测试查询结果出现跳跃，不知道为什么，感觉有点奇怪，分析看看。
--//ITPUB原始链接已经不存在，我的日记本还有记录，现在想想当时的记录思路很乱，不过这些都是猜测的过程，以前思路混乱也是正常的。
--//顺便做一些必要补充。

1.环境:
SCOTT@test01p> @ ver1
PORT_STRING          VERSION    BANNER                                                                       CON_ID
-------------------- ---------- ---------------------------------------------------------------------------- ------
IBMPC/WIN_NT64-9.1.0 12.2.0.1.0 Oracle Database 12c Enterprise Edition Release 12.2.0.1.0 - 64bit Production      0

2.测试:
SCOTT@test01p> create table t as select rownum a1 , rownum a2 ,rownum a3 from dual connect by level<=100 ;
Table created.

--//分析略.
select 'explain plan set statement_id='''||lpad(rownum,3,'0')||''''||' for select 1 from t where rownum<='||rownum||';' c80 from t;
--//把以上的输出保存一个文件执行,然后执行如下:

select STATEMENT_ID,CPU_COST,lead(cpu_cost ) over ( order by STATEMENT_ID ) N1,lead(cpu_cost ) over ( order by STATEMENT_ID )- cpu_cost N2 from (
select STATEMENT_ID,OPERATION, OPTIONS, COST, CPU_COST, IO_COST, TIME from plan_table where options='FULL');

STATEMENT_ CPU_COST    N1   N2
---------- -------- ----- ----
001            7271  7421  150
002            7421  7571  150
003            7571  7721  150
004            7721  7871  150
005            7871  8021  150
006            8021  8321  300
007            8321  8321    0
008            8321  8471  150
009            8471  8621  150
010            8621  8771  150
011            8771  8921  150
012            8921  9071  150
013            9071  9371  300
014            9371  9371    0
015            9371  9521  150
016            9521  9671  150
017            9671  9821  150
018            9821  9971  150
019            9971 10121  150
020           10121 10271  150
021           10271 10421  150
022           10421 10571  150
023           10571 10721  150
024           10721 10871  150
025           10871 18143 7272
026           18143 18293  150
027           18293 18593  300
028           18593 18593    0
029           18593 18743  150
030           18743 18893  150
031           18893 19043  150
032           19043 19193  150
033           19193 19343  150
034           19343 19493  150
035           19493 19643  150
036           19643 19793  150
037           19793 19943  150
038           19943 20093  150
039           20093 20243  150
040           20243 20393  150
041           20393 20543  150
042           20543 20693  150
043           20693 20843  150
044           20843 20993  150
045           20993 21143  150
046           21143 21293  150
047           21293 21443  150
048           21443 21593  150
049           21593 21743  150
050           21743 29014 7271
051           29014 29164  150
052           29164 29314  150
053           29314 29464  150
054           29464 29914  450
055           29914 29914    0
056           29914 29914    0
057           29914 30064  150
058           30064 30214  150
059           30214 30364  150
060           30364 30514  150
061           30514 30664  150
062           30664 30814  150
063           30814 30964  150
064           30964 31114  150
065           31114 31264  150
066           31264 31414  150
067           31414 31564  150
068           31564 31714  150
069           31714 31864  150
070           31864 32014  150
071           32014 32164  150
072           32164 32314  150
073           32314 32464  150
074           32464 32614  150
075           32614 39886 7272
076           39886 40036  150
077           40036 40186  150
078           40186 40336  150
079           40336 40486  150
080           40486 40636  150
081           40636 40786  150
082           40786 40936  150
083           40936 41086  150
084           41086 41236  150
085           41236 41386  150
086           41386 41536  150
087           41536 41686  150
088           41686 41836  150
089           41836 41986  150
090           41986 42136  150
091           42136 42286  150
092           42286 42436  150
093           42436 42586  150
094           42586 42736  150
095           42736 42886  150
096           42886 43036  150
097           43036 43186  150
098           43186 43486  300
099           43486 43486    0
100           43486
100 rows selected.
--//大于7271的部分,我前面已经解析.
--//在STATEMENT_ID=025,050,075,N2分别是7272,7271,7272.说明在statement_id=026,051,076多访问1块。
--//可以这么理解表T占4blocks,共100行，平均下来每块25行。这样当查询等于rownum<=26,51,76时出现多访问1块的情况。

--//剩下就是为什么查询条件rownum<=55,rownum<=56,rownum<=57时CPU_COST一样的,不好理解。N2出现跳跃的情况呢?

3.继续探究:
SCOTT@test01p> set feedback only
SCOTT@test01p> select 1 from t where rownum<=55 ;
         1
----------

55 rows selected.

SCOTT@test01p> set feedback 6
SCOTT@test01p> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID  g2r21fyyf3y90, child number 1
-------------------------------------
select 1 from t where rownum<=55
Plan hash value: 508354683
-------------------------------------------------------------------------------------------------------------
| Id  | Operation          | Name | Starts | E-Rows | Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |
-------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |      1 |        |     3 (100)|          |     55 |00:00:00.01 |       6 |
|*  1 |  COUNT STOPKEY     |      |      1 |        |            |          |     55 |00:00:00.01 |       6 |
|   2 |   TABLE ACCESS FULL| T    |      1 |     57 |     3   (0)| 00:00:01 |     55 |00:00:00.01 |       6 |
-------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1
   2 - SEL$1 / T@SEL$1
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter(ROWNUM<=55)
--//注意看E-Rows = 57.噢!这样明白为什么出现跳跃.是oracle估计选择率的算法非常特别造成这样的情况.
--//感觉oracle这样条件算法有点奇怪,什么可能查询条件rownum<=55,E-Rows = 57呢?

--//看看select 1 from t where rownum<=7的执行计划也可以验证:
SCOTT@test01p> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID  a8yj08mysamg1, child number 0
-------------------------------------
select 1 from t where rownum<=7
Plan hash value: 508354683
-------------------------------------------------------------------------------------------------------------
| Id  | Operation          | Name | Starts | E-Rows | Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |
-------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |      1 |        |     2 (100)|          |      7 |00:00:00.01 |       6 |
|*  1 |  COUNT STOPKEY     |      |      1 |        |            |          |      7 |00:00:00.01 |       6 |
|   2 |   TABLE ACCESS FULL| T    |      1 |      8 |     2   (0)| 00:00:01 |      7 |00:00:00.01 |       6 |
-------------------------------------------------------------------------------------------------------------
--//E-ROWS=8.这样就明白为什么我当时的算法每行成本出现跳跃的情况.其它大家可以自行验证.

4.选择率如何计算呢?
--//rownun<=N,这样的查询我看了<基于成本的Oracle优化法则>,也没有这方面的内容.
--//我试图按照区间的算法不对.
--// Selectivity =  (limit – low_value) / (high_value – low_value) + 1/num_distinct
--//(55-1)/(100-1)+1/100 = .55545454545454545454 , 不对,rownum虚拟列,这样的查询条件选择率如何确定不知道.
SCOTT@test01p> set feedback only
SCOTT@test01p> select 1 from t where a1<=55 ;
         1
----------

55 rows selected.

SCOTT@test01p> set feedback 6
SCOTT@test01p> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID  4vmjyzbu16y74, child number 0
-------------------------------------
select 1 from t where a1<=55

Plan hash value: 1601196873

--------------------------------------------------------------------------------------------------------------------
| Id  | Operation         | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time   | A-Rows |   A-Time   | Buffers |
--------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |      1 |        |       |     3 (100)|          |     55 |00:00:00.01 |      10 |
|*  1 |  TABLE ACCESS FULL| T    |      1 |     56 |   168 |     3   (0)| 00:00:01 |     55 |00:00:00.01 |      10 |
--------------------------------------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
   1 - SEL$1 / T@SEL$1
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter("A1"<=55)

--//如果查询select 1 from t where a1<=55;E-Rows=56,按照上面公式Selectivity =  (limit – low_value) / (high_value – low_value) + 1/num_distinct
--//计算的结果是正确的.
--//仅仅知道为什么出现上面的情况,不知道条件rownum<=N的选择率计算公式.
--//如果加大NUMROWS=> 1000,就不会出现前面的情况.

SCOTT@test01p> exec dbms_stats.set_table_stats(ownname=> NULL,TABNAME=>'T',NUMROWS=> 1000);
PL/SQL procedure successfully completed.

SCOTT@test01p> alter system flush shared_pool;
System altered.

--//select 'explain plan set statement_id='''||lpad(rownum,3,'0')||''''||' for select 1 from t where rownum<='||rownum||';' c80 from t;

SCOTT@test01p> SELECT *
  FROM (SELECT STATEMENT_ID,CPU_COST,lead(cpu_cost )
  OVER ( ORDER BY STATEMENT_ID ) N1,lead(cpu_cost )
  OVER ( ORDER BY STATEMENT_ID )- cpu_cost N2
  FROM ( SELECT STATEMENT_ID,
       OPERATION, OPTIONS, COST, CPU_COST, IO_COST,
       TIME FROM plan_table WHERE options =  'FULL') )
 WHERE N2 <> 150;

no rows selected