巧用函数索引性能调优-mysql教程-PHP中文网

首页

数据库

mysql教程

巧用函数索引性能调优

WBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWBOYWB

Jun 07, 2016 pm 03:28 PM

sql函数性能执行索引调优

今天调了一条SQL，执行计划中出现BITMAP CONVERSION，检查了下索引，没有位图索引

今天调了一条SQL，执行计划中出现BITMAP CONVERSION，检查了下索引，没有位图索引啊。原来是Oracle有时候会选择将B-Tree索引进行BITMAP转换来进行SQL执行，从而导致极其恶劣的执行计划。在下列计划中BITMAP CONVERSION FROM / TO ROWIDS 就是进行了位图转换后的执行计划：

SQL> SELECT *

2 FROM (SELECT A.ASSET_ID,
3 C.CLASSIFY_CODE,
4 C.CLASSIFY_NAME,
5 V.NOMINAL_VOLTAGE BASE_NOMINAL_VOLTAGE,
6 TO_CHAR(SDO_UTIL.TO_GMLGEOMETRY(A.G3E_GEOMETRY)) AS GML
7 FROM DM_ASSET A, DM_CLASSIFY C, DM_BASE_VOLTAGE V
8 WHERE A.CLASSIFY_ID = C.CLASSIFY_ID
9 AND A.BASE_VOLTAGE_ID = V.BASE_VOLTAGE_ID(+)
10 AND C.CLASSIFY_ID = '7002321'
11 AND SDO_RELATE(A.G3E_GEOMETRY,
12 SDO_GEOMETRY(3003,
13 4326,
14 NULL,
15 SDO_ELEM_INFO_ARRAY(1, 1003, 1),
16 SDO_ORDINATE_ARRAY(113.93897922622,
17 22.806658666304,
18 0,
19 114.38475977774,
20 22.806658666304,
21 0,
22 114.38475977774,
23 22.42623522295,
24 0,
25 113.93897922622,
26 22.42623522295,
27 0,
28 113.93897922622,
29 22.806658666304,
30 0)),
31 'MASK=ANYINTERACT') = 'TRUE')
32 WHERE ROWNUM 已选择500行。
已用时间: 00: 00: 06.71
执行计划
----------------------------------------------------------
Plan hash value: 1161815771
-------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 500 | 130K| 263 (1)| 00:00:04 |
|* 1 | COUNT STOPKEY | | | | | |
|* 2 | HASH JOIN RIGHT OUTER | | 501 | 130K| 263 (1)| 00:00:04 |
| 3 | TABLE ACCESS FULL | DM_BASE_VOLTAGE | 81 | 729 | 3 (0)| 00:00:01 |
| 4 | NESTED LOOPS | | 501 | 64629 | 259 (1)| 00:00:04 |
| 5 | TABLE ACCESS BY INDEX ROWID | DM_CLASSIFY | 1 | 23 | 2 (0)| 00:00:01 |
|* 6 | INDEX UNIQUE SCAN | IDX_DM_CLASSIFY_ID | 1 | | 1 (0)| 00:00:01 |
| 7 | TABLE ACCESS BY INDEX ROWID | DM_ASSET | 501 | 53106 | 259 (1)| 00:00:04 |
| 8 | BITMAP CONVERSION TO ROWIDS | | | | | |
| 9 | BITMAP AND | | | | | |
| 10 | BITMAP CONVERSION FROM ROWIDS| | | | | |
| 11 | SORT ORDER BY | | | | | |
|* 12 | DOMAIN INDEX | IDX_DM_ASSET_SPL | 50100 | | 0 (0)| 00:00:01 |
| 13 | BITMAP CONVERSION FROM ROWIDS| | | | | |
|* 14 | INDEX RANGE SCAN | IDX_ASSET_CLASSIFY_ID | 50100 | | 135 (1)| 00:00:02 |
-------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM 2 - access("A"."BASE_VOLTAGE_ID"="V"."BASE_VOLTAGE_ID"(+))
6 - access("C"."CLASSIFY_ID"='7002321')
12 - access("MDSYS"."SDO_RELATE"("A"."G3E_GEOMETRY","MDSYS"."SDO_GEOMETRY"(3003,4326,NULL,"SDO_ELEM
_INFO_ARRAY"(1,1003,1),"SDO_ORDINATE_ARRAY"(113.93897922622,22.806658666304,0,114.38475977774,22.8066
58666304,0,114.38475977774,22.42623522295,0,113.93897922622,22.42623522295,0,113.93897922622,22.80665
8666304,0)),'MASK=ANYINTERACT')='TRUE')
14 - access("A"."CLASSIFY_ID"='7002321')
统计信息
----------------------------------------------------------
76096 recursive calls
6004 db block gets
78944 consistent gets
0 physical reads
0 redo size
100625 bytes sent via SQL*Net to client
748 bytes received via SQL*Net from client
35 SQL*Net roundtrips to/from client
3 sorts (memory)
0 sorts (disk)
500 rows processed

到网上找了下，有两种解决方法，一种是将_b_tree_bitmap_plans设置为false，另一种是将选择性差的索引去掉。

尝试用第一种方法：

SQL> alter session set "_b_tree_bitmap_plans"=false;
SQL> SELECT *
2 FROM (SELECT A.ASSET_ID,
3 C.CLASSIFY_CODE,
4 C.CLASSIFY_NAME,
5 V.NOMINAL_VOLTAGE BASE_NOMINAL_VOLTAGE,
6 TO_CHAR(SDO_UTIL.TO_GMLGEOMETRY(A.G3E_GEOMETRY)) AS GML
7 FROM DM_ASSET A, DM_CLASSIFY C, DM_BASE_VOLTAGE V
8 WHERE A.CLASSIFY_ID = C.CLASSIFY_ID
9 AND A.BASE_VOLTAGE_ID = V.BASE_VOLTAGE_ID(+)
10 AND C.CLASSIFY_ID = '7002321'
11 AND SDO_RELATE(A.G3E_GEOMETRY,
12 SDO_GEOMETRY(3003,
13 4326,
14 NULL,
15 SDO_ELEM_INFO_ARRAY(1, 1003, 1),
16 SDO_ORDINATE_ARRAY(113.93897922622,
17 22.806658666304,
18 0,
19 114.38475977774,
20 22.806658666304,
21 0,
22 114.38475977774,
23 22.42623522295,
24 0,
25 113.93897922622,
26 22.42623522295,
27 0,
28 113.93897922622,
29 22.806658666304,
30 0)),
31 'MASK=ANYINTERACT') = 'TRUE')
32 WHERE ROWNUM 已选择500行。
已用时间: 00: 00: 06.51
执行计划
----------------------------------------------------------
Plan hash value: 1161815771
-------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 500 | 130K| 263 (1)| 00:00:04 |
|* 1 | COUNT STOPKEY | | | | | |
|* 2 | HASH JOIN RIGHT OUTER | | 501 | 130K| 263 (1)| 00:00:04 |
| 3 | TABLE ACCESS FULL | DM_BASE_VOLTAGE | 81 | 729 | 3 (0)| 00:00:01 |
| 4 | NESTED LOOPS | | 501 | 64629 | 259 (1)| 00:00:04 |
| 5 | TABLE ACCESS BY INDEX ROWID | DM_CLASSIFY | 1 | 23 | 2 (0)| 00:00:01 |
|* 6 | INDEX UNIQUE SCAN | IDX_DM_CLASSIFY_ID | 1 | | 1 (0)| 00:00:01 |
| 7 | TABLE ACCESS BY INDEX ROWID | DM_ASSET | 501 | 53106 | 259 (1)| 00:00:04 |
| 8 | BITMAP CONVERSION TO ROWIDS | | | | | |
| 9 | BITMAP AND | | | | | |
| 10 | BITMAP CONVERSION FROM ROWIDS| | | | | |
| 11 | SORT ORDER BY | | | | | |
|* 12 | DOMAIN INDEX | IDX_DM_ASSET_SPL | 50100 | | 0 (0)| 00:00:01 |
| 13 | BITMAP CONVERSION FROM ROWIDS| | | | | |
|* 14 | INDEX RANGE SCAN | IDX_ASSET_CLASSIFY_ID | 50100 | | 135 (1)| 00:00:02 |
-------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM 2 - access("A"."BASE_VOLTAGE_ID"="V"."BASE_VOLTAGE_ID"(+))
6 - access("C"."CLASSIFY_ID"='7002321')
12 - access("MDSYS"."SDO_RELATE"("A"."G3E_GEOMETRY","MDSYS"."SDO_GEOMETRY"(3003,4326,NULL,"SDO_ELEM
_INFO_ARRAY"(1,1003,1),"SDO_ORDINATE_ARRAY"(113.93897922622,22.806658666304,0,114.38475977774,22.8066
58666304,0,114.38475977774,22.42623522295,0,113.93897922622,22.42623522295,0,113.93897922622,22.80665
8666304,0)),'MASK=ANYINTERACT')='TRUE')
14 - access("A"."CLASSIFY_ID"='7002321')
统计信息
----------------------------------------------------------
76096 recursive calls
6002 db block gets
78944 consistent gets
0 physical reads
0 redo size
100625 bytes sent via SQL*Net to client
748 bytes received via SQL*Net from client
35 SQL*Net roundtrips to/from client
3 sorts (memory)
0 sorts (disk)
500 rows processed

DM_ASSET有7百多万的数据，而CLASSIFY_ID的值只有39个，其中38个值可以走索引，一个值的数量非常多，典型的数据不均匀。
select CLASSIFY_ID,count(1) from DM_ASSET group by CLASSIFY_ID order by 2;
7002386 1
7002369 3
7002381 13
7002513 16
7002349 18
7002333 36
7002474 166
7002502 276
7002345 1131
7002344 1423
7002382 1709
7002359 1791
7002510 6340
7002374 6684
7002358 7286
7002361 8750
7002379 11189
7002340 12622
7002473 12717
7002348 14901
7002360 15722
7002335 17939
7002336 21492
7002500 28304
7002363 42883
7002343 49865
7002472 51327
7002321 76098
7002373 89485
7002515 110765
7002378 128512
7002380 147776
7002499 235166
7002370 271190
7002501 370439
7002398 456259
7002496 830986
7002401 4361079

尝试用第二种方法：删除CLASSIFY_ID上的索引。
SQL> SELECT *
2 FROM (SELECT A.ASSET_ID,
3 C.CLASSIFY_CODE,
4 C.CLASSIFY_NAME,
5 V.NOMINAL_VOLTAGE BASE_NOMINAL_VOLTAGE,
6 TO_CHAR(SDO_UTIL.TO_GMLGEOMETRY(A.G3E_GEOMETRY)) AS GML
7 FROM DM_ASSET A, DM_CLASSIFY C, DM_BASE_VOLTAGE V
8 WHERE A.CLASSIFY_ID = C.CLASSIFY_ID
9 AND A.BASE_VOLTAGE_ID = V.BASE_VOLTAGE_ID(+)
10 AND C.CLASSIFY_ID = '7002321'
11 AND SDO_RELATE(A.G3E_GEOMETRY,
12 SDO_GEOMETRY(3003,
13 4326,
14 NULL,
15 SDO_ELEM_INFO_ARRAY(1, 1003, 1),
16 SDO_ORDINATE_ARRAY(113.93897922622,
17 22.806658666304,
18 0,
19 114.38475977774,
20 22.806658666304,
21 0,
22 114.38475977774,
23 22.42623522295,
24 0,
25 113.93897922622,
26 22.42623522295,
27 0,
28 113.93897922622,
29 22.806658666304,
30 0)),
31 'MASK=ANYINTERACT') = 'TRUE')
32 WHERE ROWNUM 已选择500行。
已用时间: 00: 00: 00.50
执行计划
----------------------------------------------------------
Plan hash value: 4025821404
-----------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 500 | 130K| 10563 (1)| 00:02:07 |
|* 1 | COUNT STOPKEY | | | | | |
|* 2 | HASH JOIN RIGHT OUTER | | 501 | 130K| 10563 (1)| 00:02:07 |
| 3 | TABLE ACCESS FULL | DM_BASE_VOLTAGE | 81 | 729 | 3 (0)| 00:00:01 |
| 4 | NESTED LOOPS | | 501 | 64629 | 10560 (1)| 00:02:07 |
| 5 | TABLE ACCESS BY INDEX ROWID| DM_CLASSIFY | 1 | 23 | 2 (0)| 00:00:01 |
|* 6 | INDEX UNIQUE SCAN | IDX_DM_CLASSIFY_ID | 1 | | 1 (0)| 00:00:01 |
|* 7 | TABLE ACCESS BY INDEX ROWID| DM_ASSET | 501 | 53106 | 10560 (1)| 00:02:07 |
|* 8 | DOMAIN INDEX | IDX_DM_ASSET_SPL | | | 0 (0)| 00:00:01 |
-----------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM 2 - access("A"."BASE_VOLTAGE_ID"="V"."BASE_VOLTAGE_ID"(+))
6 - access("C"."CLASSIFY_ID"='7002321')
7 - filter("A"."CLASSIFY_ID"='7002321')
8 - access("MDSYS"."SDO_RELATE"("A"."G3E_GEOMETRY","MDSYS"."SDO_GEOMETRY"(3003,4326,NULL,"
SDO_ELEM_INFO_ARRAY"(1,1003,1),"SDO_ORDINATE_ARRAY"(113.93897922622,22.806658666304,0,114.384
75977774,22.806658666304,0,114.38475977774,22.42623522295,0,113.93897922622,22.42623522295,0,
113.93897922622,22.806658666304,0)),'MASK=ANYINTERACT')='TRUE')
统计信息
----------------------------------------------------------
3965 recursive calls
6006 db block gets
30388 consistent gets
3 physical reads
340 redo size
94355 bytes sent via SQL*Net to client
748 bytes received via SQL*Net from client
35 SQL*Net roundtrips to/from client
13 sorts (memory)
0 sorts (disk)
500 rows processed

问题又来了，删除索引后，这个SQL是快了，但原本根据CLASSIFY_ID走索引的语句性能出了问题。曾经在itpub上看到newid这么用，建一个函数索引：
CREATE INDEX IDX_ASSET_CLASSIFY_ID ON DM_ASSET(CASE WHEN CLASSIFY_ID !='7002401' THEN CLASSIFY_ID ELSE NULL END) nologging;
1. 上面的SQL执行也是非常快。
2. 根据CLASSIFY_ID的代码需要改一下
if(CLASSIFY_ID = '7002401') {--如果是这个值，则走全部扫描
select * from DM_ASSET where CLASSIFY_ID = '7002321';
} else {--如果是这个值，则走索引
select * from DM_ASSET where (CASE WHEN CLASSIFY_ID !='7002401'
THEN CLASSIFY_ID ELSE NULL END)='7002401'
}

声明

本文内容由网友自发贡献，版权归原作者所有，本站不承担相应法律责任。如您发现有涉嫌抄袭侵权的内容，请联系admin@php.cn

您什么时候应该使用复合索引与多个单列索引？Apr 11, 2025 am 12:06 AM

在数据库优化中，应根据查询需求选择索引策略：1.当查询涉及多个列且条件顺序固定时，使用复合索引；2.当查询涉及多个列但条件顺序不固定时，使用多个单列索引。复合索引适用于优化多列查询，单列索引则适合单列查询。

如何识别和优化MySQL中的慢速查询？（慢查询日志，performance_schema）Apr 10, 2025 am 09:36 AM

要优化MySQL慢查询，需使用slowquerylog和performance_schema：1.启用slowquerylog并设置阈值，记录慢查询；2.利用performance_schema分析查询执行细节，找出性能瓶颈并优化。

MySQL和SQL：开发人员的基本技能Apr 10, 2025 am 09:30 AM

MySQL和SQL是开发者必备技能。1.MySQL是开源的关系型数据库管理系统，SQL是用于管理和操作数据库的标准语言。2.MySQL通过高效的数据存储和检索功能支持多种存储引擎，SQL通过简单语句完成复杂数据操作。3.使用示例包括基本查询和高级查询，如按条件过滤和排序。4.常见错误包括语法错误和性能问题，可通过检查SQL语句和使用EXPLAIN命令优化。5.性能优化技巧包括使用索引、避免全表扫描、优化JOIN操作和提升代码可读性。

描述MySQL异步主奴隶复制过程。Apr 10, 2025 am 09:30 AM

MySQL异步主从复制通过binlog实现数据同步，提升读性能和高可用性。1)主服务器记录变更到binlog；2)从服务器通过I/O线程读取binlog；3)从服务器的SQL线程应用binlog同步数据。

mysql：简单的概念，用于轻松学习Apr 10, 2025 am 09:29 AM

MySQL是一个开源的关系型数据库管理系统。1）创建数据库和表：使用CREATEDATABASE和CREATETABLE命令。2）基本操作：INSERT、UPDATE、DELETE和SELECT。3）高级操作：JOIN、子查询和事务处理。4）调试技巧：检查语法、数据类型和权限。5）优化建议：使用索引、避免SELECT*和使用事务。

MySQL：数据库的用户友好介绍Apr 10, 2025 am 09:27 AM

MySQL的安装和基本操作包括：1.下载并安装MySQL，设置根用户密码；2.使用SQL命令创建数据库和表，如CREATEDATABASE和CREATETABLE；3.执行CRUD操作，使用INSERT,SELECT,UPDATE,DELETE命令；4.创建索引和存储过程以优化性能和实现复杂逻辑。通过这些步骤，你可以从零开始构建和管理MySQL数据库。

InnoDB缓冲池如何工作，为什么对性能至关重要？Apr 09, 2025 am 12:12 AM

InnoDBBufferPool通过将数据和索引页加载到内存中来提升MySQL数据库的性能。1)数据页加载到BufferPool中，减少磁盘I/O。2)脏页被标记并定期刷新到磁盘。3)LRU算法管理数据页淘汰。4)预读机制提前加载可能需要的数据页。

MySQL：初学者的数据管理易用性Apr 09, 2025 am 12:07 AM

MySQL适合初学者使用，因为它安装简单、功能强大且易于管理数据。1.安装和配置简单，适用于多种操作系统。2.支持基本操作如创建数据库和表、插入、查询、更新和删除数据。3.提供高级功能如JOIN操作和子查询。4.可以通过索引、查询优化和分表分区来提升性能。5.支持备份、恢复和安全措施，确保数据的安全和一致性。

See all articles