« Oracle9i新特性：iSQLPLUS | Blog首页 | CBO对于Oracle SQL执行计划的影响(之二) »
CBO对于Oracle SQL执行计划的影响(之一)
作者：eygle | 【转载请注出处】|【云和恩墨 领先的zData数据库一体机 | zCloud PaaS云管平台 | SQM SQL审核平台 | ZDBM 数据库备份一体机】
链接：https://www.eygle.com/archives/2004/10/how_cbo_effect_sql_explain01.html
链接：https://www.eygle.com/archives/2004/10/how_cbo_effect_sql_explain01.html
1. 原始SQL语句
这个SQL语句是一个动态查询语句的一部分，该查询根据不同条件生成不同的SQL语句。
本例为查询2003年以来的入库单据，很少的数据。
SELECT "SP_TRANS"."TRANS_NO",
"SP_TRANS"."TRANS_TYPE",
"SP_TRANS"."STORE_NO",
"SP_TRANS"."BILL_NO",
"SP_TRANS"."TRANSDATE",
"SP_TRANS"."MANAGER_ID",
"SP_TRANS"."REMARK",
"SP_TRANS"."STATE",
"SP_TRANS_SUB"."TRANS_NO",
"SP_TRANS_SUB"."ITEM_CODE",
"SP_TRANS_SUB"."COUNTRY",
"SP_TRANS_SUB"."QTY",
"SP_TRANS_SUB"."PRICE",
"SP_TRANS_SUB"."TOTAL",
"SP_CHK"."CHK_NO",
"SP_CHK"."RECEIVE_NO",
"SP_CHK"."CHECKER",
"SP_CHK_SUB"."CHK_NO",
"SP_CHK_SUB"."ITEM_CODE",
"SP_CHK_SUB"."COUNTRY",
"SP_CHK_SUB"."PLAN_NO",
"SP_CHK_SUB"."PLAN_LINE",
"SP_CHK_SUB"."QTY_CHECKOUT",
"SP_CHK_SUB"."NOW_QTY",
"SP_RECEIVE"."RECEIVE_NO",
"SP_RECEIVE"."VENDOR_NAME",
"SP_RECEIVE"."BUYER",
"SP_RECEIVE_SUB"."RECEIVE_NO",
"SP_RECEIVE_SUB"."PLAN_NO",
"SP_RECEIVE_SUB"."PLAN_LINE",
"SP_RECEIVE_SUB"."ITEM_NAME",
"SP_RECEIVE_SUB"."COUNTRY",
"SP_ITEM"."ITEM_CODE",
"SP_ITEM"."CHART_ID",
"SP_ITEM"."SPECIFICATION"
FROM "SP_TRANS",
"SP_TRANS_SUB",
"SP_CHK",
"SP_CHK_SUB",
"SP_RECEIVE",
"SP_RECEIVE_SUB",
"SP_ITEM"
WHERE ( "SP_TRANS_SUB"."TRANS_NO" = "SP_TRANS"."TRANS_NO" ) and
("SP_TRANS"."BILL_NO" = "SP_CHK"."CHK_NO") and
( "SP_CHK_SUB"."CHK_NO" = "SP_CHK"."CHK_NO" ) and
( "SP_CHK"."RECEIVE_NO" = "SP_RECEIVE"."RECEIVE_NO" ) and
( "SP_CHK"."STATE" = 15 ) and
( "SP_RECEIVE_SUB"."RECEIVE_NO" = "SP_RECEIVE"."RECEIVE_NO" ) and
( "SP_TRANS_SUB"."ITEM_CODE" = "SP_ITEM"."ITEM_CODE" ) and
( "SP_TRANS_SUB"."ITEM_CODE" = "SP_CHK_SUB"."ITEM_CODE" ) and
( "SP_CHK_SUB"."ITEM_CODE" = "SP_RECEIVE_SUB"."ITEM_CODE" ) and
( "SP_CHK_SUB"."COUNTRY" = "SP_TRANS_SUB"."COUNTRY" ) and
( "SP_CHK_SUB"."COUNTRY" = "SP_RECEIVE_SUB"."COUNTRY" ) and
( "SP_CHK_SUB"."PLAN_NO" = "SP_RECEIVE_SUB"."PLAN_NO" ) and
( "SP_CHK_SUB"."PLAN_LINE" = "SP_RECEIVE_SUB"."PLAN_LINE" ) and
(to_char("SP_TRANS"."TRANSDATE" ,'YYYY-MM-DD') >='2003-01-01')
/
|
2. 执行计划
我们的数据库使用dbms_stats.gather_schema_stats分析过，具有足够及时的所有数据，然而在CBO的执行计划下，优化器选择了完全
不同的执行计划.
a. no hints
这是未加任何提示时，Oralce选择的执行路径，在实际程序中，用户说死掉了，通过执行计划我们知道，不是死掉了，是慢！！！
|
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2057 Card=1 Bytes=288)
1 0 NESTED LOOPS (Cost=2057 Card=1 Bytes=288)
2 1 NESTED LOOPS (Cost=2056 Card=1 Bytes=256)
3 2 NESTED LOOPS (Cost=2054 Card=1 Bytes=219)
4 3 NESTED LOOPS (Cost=2053 Card=1 Bytes=178)
5 4 NESTED LOOPS (Cost=2009 Card=1 Bytes=131)
6 5 MERGE JOIN (Cost=2008 Card=1 Bytes=100)
7 6 SORT (JOIN) (Cost=950 Card=36412 Bytes=1747776)
8 7 TABLE ACCESS (FULL) OF 'SP_CHK_SUB' (Cost=59 Card=36412 Bytes=1747776)
9 6 SORT (JOIN) (Cost=1058 Card=36730 Bytes=1909960)
10 9 TABLE ACCESS (FULL) OF 'SP_RECEIVE_SUB' (Cost=89 Card=36730 Bytes=1909960)
11 5 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK' (Cost=1 Card=3870 Bytes=119970)
12 11 INDEX (UNIQUE SCAN) OF 'PK_SP_CHK' (UNIQUE)
13 4 TABLE ACCESS (FULL) OF 'SP_TRANS' (Cost=44 Card=1717 Bytes=80699)
14 3 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE' (Cost=1 Card=7816 Bytes=320456)
15 14 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE' (UNIQUE)
16 2 TABLE ACCESS (BY INDEX ROWID) OF 'SP_TRANS_SUB' (Cost=2 Card=136371 Bytes=5045727)
17 16 INDEX (UNIQUE SCAN) OF 'PK_SP_TRANS_SUB' (UNIQUE) (Cost=1 Card=136371)
18 1 TABLE ACCESS (BY INDEX ROWID) OF 'SP_ITEM' (Cost=1 Card=29763 Bytes=952416)
19 18 INDEX (UNIQUE SCAN) OF 'SYS_C0012193' (UNIQUE)
|
用足够的耐心，我们得到了该计划的执行结果。
|
SQL> SELECT "SP_TRANS"."TRANS_NO",
2 "SP_TRANS"."TRANS_TYPE",
3 "SP_TRANS"."STORE_NO",
4 "SP_TRANS"."BILL_NO",
5 "SP_TRANS"."TRANSDATE",
6 "SP_TRANS"."MANAGER_ID",
7 "SP_TRANS"."REMARK",
8 "SP_TRANS"."STATE",
9 "SP_TRANS_SUB"."TRANS_NO",
10 "SP_TRANS_SUB"."ITEM_CODE",
11 "SP_TRANS_SUB"."COUNTRY",
12 "SP_TRANS_SUB"."QTY",
13 "SP_TRANS_SUB"."PRICE",
14 "SP_TRANS_SUB"."TOTAL",
15 "SP_CHK"."CHK_NO",
16 "SP_CHK"."RECEIVE_NO",
17 "SP_CHK"."CHECKER",
18 "SP_CHK_SUB"."CHK_NO",
19 "SP_CHK_SUB"."ITEM_CODE",
20 "SP_CHK_SUB"."COUNTRY",
21 "SP_CHK_SUB"."PLAN_NO",
22 "SP_CHK_SUB"."PLAN_LINE",
23 "SP_CHK_SUB"."QTY_CHECKOUT",
24 "SP_CHK_SUB"."NOW_QTY",
25 "SP_RECEIVE"."RECEIVE_NO",
26 "SP_RECEIVE"."VENDOR_NAME",
27 "SP_RECEIVE"."BUYER",
28 "SP_RECEIVE_SUB"."RECEIVE_NO",
29 "SP_RECEIVE_SUB"."PLAN_NO",
30 "SP_RECEIVE_SUB"."PLAN_LINE",
31 "SP_RECEIVE_SUB"."ITEM_NAME",
32 "SP_RECEIVE_SUB"."COUNTRY",
33 "SP_ITEM"."ITEM_CODE",
34 "SP_ITEM"."CHART_ID",
35 "SP_ITEM"."SPECIFICATION"
36 FROM "SP_TRANS",
37 "SP_TRANS_SUB",
38 "SP_CHK",
39 "SP_CHK_SUB",
40 "SP_RECEIVE",
41 "SP_RECEIVE_SUB",
42 "SP_ITEM"
43 WHERE ( "SP_TRANS_SUB"."TRANS_NO" = "SP_TRANS"."TRANS_NO" ) and
44 ( "SP_TRANS"."BILL_NO" = "SP_CHK"."CHK_NO") and
45 ( "SP_CHK_SUB"."CHK_NO" = "SP_CHK"."CHK_NO" ) and
46 ( "SP_CHK"."RECEIVE_NO" = "SP_RECEIVE"."RECEIVE_NO" ) and
47 ( "SP_CHK"."STATE" = 15 ) and
48 ( "SP_RECEIVE_SUB"."RECEIVE_NO" = "SP_RECEIVE"."RECEIVE_NO" ) and
49 ( "SP_TRANS_SUB"."ITEM_CODE" = "SP_ITEM"."ITEM_CODE" ) and
50 ( "SP_TRANS_SUB"."ITEM_CODE" = "SP_CHK_SUB"."ITEM_CODE" ) and
51 ( "SP_CHK_SUB"."ITEM_CODE" = "SP_RECEIVE_SUB"."ITEM_CODE" ) and
52 ( "SP_CHK_SUB"."COUNTRY" = "SP_TRANS_SUB"."COUNTRY" ) and
53 ( "SP_CHK_SUB"."COUNTRY" = "SP_RECEIVE_SUB"."COUNTRY" ) and
54 ( "SP_CHK_SUB"."PLAN_NO" = "SP_RECEIVE_SUB"."PLAN_NO" ) and
55 ( "SP_CHK_SUB"."PLAN_LINE" = "SP_RECEIVE_SUB"."PLAN_LINE" ) and
56 (to_char("SP_TRANS"."TRANSDATE" ,'YYYY-MM-DD') >='2003-01-01')
57 /
130 rows selected.
Elapsed: 00: 29: 1785.47
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2057 Card=1 Bytes=288)
1 0 NESTED LOOPS (Cost=2057 Card=1 Bytes=288)
2 1 NESTED LOOPS (Cost=2056 Card=1 Bytes=256)
3 2 NESTED LOOPS (Cost=2054 Card=1 Bytes=219)
4 3 NESTED LOOPS (Cost=2053 Card=1 Bytes=178)
5 4 NESTED LOOPS (Cost=2009 Card=1 Bytes=131)
6 5 MERGE JOIN (Cost=2008 Card=1 Bytes=100)
7 6 SORT (JOIN) (Cost=950 Card=36412 Bytes=1747776)
8 7 TABLE ACCESS (FULL) OF 'SP_CHK_SUB' (Cost=59 Card=36412 Bytes=1747776)
9 6 SORT (JOIN) (Cost=1058 Card=36730 Bytes=1909960)
10 9 TABLE ACCESS (FULL) OF 'SP_RECEIVE_SUB' (Cost=89 Card=36730 Bytes=1909960)
11 5 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK' (Cost=1 Card=3870 Bytes=119970)
12 11 INDEX (UNIQUE SCAN) OF 'PK_SP_CHK' (UNIQUE)
13 4 TABLE ACCESS (FULL) OF 'SP_TRANS' (Cost=44 Card=1717 Bytes=80699)
14 3 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE' (Cost=1 Card=7816 Bytes=320456)
15 14 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE' (UNIQUE)
16 2 TABLE ACCESS (BY INDEX ROWID) OF 'SP_TRANS_SUB' (Cost=2 Card=136371 Bytes=5045727)
17 16 INDEX (UNIQUE SCAN) OF 'PK_SP_TRANS_SUB' (UNIQUE) (Cost=1 Card=136371)
18 1 TABLE ACCESS (BY INDEX ROWID) OF 'SP_ITEM' (Cost=1 Card=29763 Bytes=952416)
19 18 INDEX (UNIQUE SCAN) OF 'SYS_C0012193' (UNIQUE)
Statistics
----------------------------------------------------------
16 recursive calls
186307 db block gets
10685361 consistent gets
2329 physical reads
0 redo size
38486 bytes sent via SQL*Net to client
1117 bytes received via SQL*Net from client
10 SQL*Net roundtrips to/from client
7 sorts (memory)
2 sorts (disk)
130 rows processed
|
可以看到，该执行计划消耗了大量的资源以及时间，这种情况是无法忍受的。
b. rule
在RBO条件下，该语句是执行很快的
加入rule提示，我们得到以下执行计划：
|
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=HINT: RULE
1 0 NESTED LOOPS
2 1 NESTED LOOPS
3 2 NESTED LOOPS
4 3 NESTED LOOPS
5 4 NESTED LOOPS
6 5 NESTED LOOPS
7 6 TABLE ACCESS (FULL) OF 'SP_TRANS_SUB'
8 6 TABLE ACCESS (BY INDEX ROWID) OF 'SP_ITEM'
9 8 INDEX (UNIQUE SCAN) OF 'SYS_C0012193' (UNIQUE)
10 5 TABLE ACCESS (BY INDEX ROWID) OF 'SP_TRANS'
11 10 INDEX (UNIQUE SCAN) OF 'PK_HSP_TRANS' (UNIQUE)
12 4 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK'
13 12 INDEX (UNIQUE SCAN) OF 'PK_SP_CHK' (UNIQUE)
14 3 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE'
15 14 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE' (UNIQUE)
16 2 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK_SUB'
17 16 INDEX (RANGE SCAN) OF 'IDX_CHK_SUB_ITEM_CODE' (NON-UNIQUE)
18 1 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE_SUB'
19 18 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE_SUB' (UNIQUE)
|
执行该计划，我们得到以下输出：
|
SQL>@sql
130 rows selected.
Elapsed: 00: 00: 12.17
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=HINT: RULE
1 0 NESTED LOOPS
2 1 NESTED LOOPS
3 2 NESTED LOOPS
4 3 NESTED LOOPS
5 4 NESTED LOOPS
6 5 NESTED LOOPS
7 6 TABLE ACCESS (FULL) OF 'SP_TRANS_SUB'
8 6 TABLE ACCESS (BY INDEX ROWID) OF 'SP_ITEM'
9 8 INDEX (UNIQUE SCAN) OF 'SYS_C0012193' (UNIQUE)
10 5 TABLE ACCESS (BY INDEX ROWID) OF 'SP_TRANS'
11 10 INDEX (UNIQUE SCAN) OF 'PK_HSP_TRANS' (UNIQUE)
12 4 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK'
13 12 INDEX (UNIQUE SCAN) OF 'PK_SP_CHK' (UNIQUE)
14 3 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE'
15 14 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE' (UNIQUE)
16 2 TABLE ACCESS (BY INDEX ROWID) OF 'SP_CHK_SUB'
17 16 INDEX (RANGE SCAN) OF 'IDX_CHK_SUB_ITEM_CODE' (NON-UNIQUE)
18 1 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE_SUB'
19 18 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE_SUB' (UNIQUE)
Statistics
----------------------------------------------------------
0 recursive calls
6 db block gets
829182 consistent gets
0 physical reads
0 redo size
37383 bytes sent via SQL*Net to client
1127 bytes received via SQL*Net from client
10 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
130 rows processed
SQL>
|
c. ordered
然后我想起了Ordered提示
使用该提示的执行计划如下：
|
SQL>@sql
已选择130行。
已用时间: 00: 00: 05.67
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=3284 Card=1 Bytes=288)
1 0 NESTED LOOPS (Cost=3284 Card=1 Bytes=288)
2 1 NESTED LOOPS (Cost=3283 Card=1 Bytes=256)
3 2 MERGE JOIN (Cost=3282 Card=1 Bytes=204)
4 3 SORT (JOIN) (Cost=2333 Card=6823 Bytes=1064388)
5 4 HASH JOIN (Cost=1848 Card=6823 Bytes=1064388)
6 5 HASH JOIN (Cost=216 Card=1717 Bytes=204323)
7 6 HASH JOIN (Cost=96 Card=1717 Bytes=133926)
8 7 TABLE ACCESS (FULL) OF 'SP_TRANS' (Cost=44 Card=1717 Bytes=80699)
9 7 TABLE ACCESS (FULL) OF 'SP_CHK' (Cost=13 Card=3870 Bytes=119970)
10 6 TABLE ACCESS (FULL) OF 'SP_RECEIVE' (Cost=17 Card=7816 Bytes=320456)
11 5 TABLE ACCESS (FULL) OF 'SP_TRANS_SUB' (Cost=155 Card=136371 Bytes=5045727)
12 3 SORT (JOIN) (Cost=950 Card=36412 Bytes=1747776)
13 12 TABLE ACCESS (FULL) OF 'SP_CHK_SUB' (Cost=59 Card=36412 Bytes=1747776)
14 2 TABLE ACCESS (BY INDEX ROWID) OF 'SP_RECEIVE_SUB' (Cost=1 Card=36730 Bytes=1909960)
15 14 INDEX (UNIQUE SCAN) OF 'PK_SP_RECEIVE_SUB' (UNIQUE)
16 1 TABLE ACCESS (BY INDEX ROWID) OF 'SP_ITEM' (Cost=1 Card=29763 Bytes=952416)
17 16 INDEX (UNIQUE SCAN) OF 'SYS_C0012193' (UNIQUE)
Statistics
----------------------------------------------------------
8 recursive calls
88 db block gets
2667 consistent gets
1093 physical reads
0 redo size
37285 bytes sent via SQL*Net to client
1109 bytes received via SQL*Net from client
10 SQL*Net roundtrips to/from client
8 sorts (memory)
1 sorts (disk)
130 rows processed
SQL>
|
很幸运，Ordered提示使Oracle选择了较好的执行计划。
所以会产生这样的效果，是因为在CBO的执行计划中，对于7张数据表，Oracle需要计算7！(5040)个连接顺序，然后比较各个顺序的
成本，最后选择成本较低的执行计划。
显然，在这一判断上耗费了大量的时间。当我们使用ordered hints的时候，Oracle就不需要这一计算步骤，它只需要使用我们指定的
顺序，然后快速的给出结果。然后问题迎刃而解。
历史上的今天...
>> 2021-10-30文章:
>> 2011-10-30文章:
>> 2009-10-30文章:
>> 2008-10-30文章:
>> 2007-10-30文章:
>> 2006-10-30文章:
>> 2005-10-30文章:
By eygle on 2004-10-30 09:25 | Comments (0) | SQL.PLSQL | 78 |