记一次SQL性能优化,查询时间从4000ms优化到200ms.
以下这句SQL是从PLM中获取代办工作流的。没优化前SQL语句执行一次大概4000ms（4秒）。
select ch.change_number changeNumber, f.text changeDetail, u1.last_name creator,
l.value status, to_char(create_date,'yyyy-mm-dd hh24:mi:ss') createDate,
so.signoff_status type
from agile.change ch
inner join agile.workflow_process wp on (ch.id=wp.change_id and ch.status=wp.state and wp.changed_by is null)
inner join agile.signoff so on wp.id=so.process_id and so.signoff_status in (0,4) and so.required in(1,5)
inner join agile.agileuser u on u.id=so.user_assigned
inner join agile.langtable l on (ch.status=l.id and l.type=4450 and l.langid=4)
inner join agile.agileuser u1 on ch.originator=u1.id
left join agile.agile_flex f on f.id=ch.id and f.attid=2017
where (ch.subclass in (2473549,2473495,1,2475794,2473579,2474897,2473519,2480885,2479577,2473531,2485198,2479783)
or (ch.subclass = 2478248 and l.value = '审批'))
and u.email ='zhangsan@kedacom.com'
使用autotrace分析sql
set autotrace on
set timing on
分析结果如下：
Elapsed: 00:00:00.008
Plan hash value: 1883359798 ------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 24 | 5280 | 954 (1)| 00:00:12 |
| 1 | NESTED LOOPS OUTER | | 24 | 5280 | 954 (1)| 00:00:12 |
| 2 | NESTED LOOPS | | 22 | 3564 | 921 (1)| 00:00:12 |
| 3 | NESTED LOOPS | | 22 | 3234 | 910 (1)| 00:00:11 |
| 4 | NESTED LOOPS | | 56 | 6328 | 882 (1)| 00:00:11 |
| 5 | NESTED LOOPS | | 156 | 11076 | 804 (1)| 00:00:10 |
| 6 | NESTED LOOPS | | 212 | 10176 | 592 (1)| 00:00:08 |
|* 7 | TABLE ACCESS FULL | AGILEUSER | 1 | 29 | 168 (1)| 00:00:03 |
|* 8 | TABLE ACCESS BY INDEX ROWID| SIGNOFF | 210 | 3990 | 424 (0)| 00:00:06 |
|* 9 | INDEX RANGE SCAN | SIGNOFF_IDX4 | 1120 | | 3 (0)| 00:00:01 |
|* 10 | TABLE ACCESS BY INDEX ROWID | WORKFLOW_PROCESS | 1 | 23 | 1 (0)| 00:00:01 |
|* 11 | INDEX UNIQUE SCAN | WF_PROCESS_PK | 1 | | 1 (0)| 00:00:01 |
|* 12 | TABLE ACCESS BY INDEX ROWID | CHANGE | 1 | 42 | 1 (0)| 00:00:01 |
|* 13 | INDEX UNIQUE SCAN | CHANGE_PK | 1 | | 1 (0)| 00:00:01 |
|* 14 | TABLE ACCESS BY INDEX ROWID | LANGTABLE | 1 | 34 | 1 (0)| 00:00:01 |
|* 15 | INDEX UNIQUE SCAN | LANGTABLE_PK | 1 | | 1 (0)| 00:00:01 |
| 16 | TABLE ACCESS BY INDEX ROWID | AGILEUSER | 1 | 15 | 1 (0)| 00:00:01 |
|* 17 | INDEX UNIQUE SCAN | AGILEUSER_PK | 1 | | 1 (0)| 00:00:01 |
| 18 | TABLE ACCESS BY INDEX ROWID | AGILE_FLEX | 1 | 58 | 2 (0)| 00:00:01 |
|* 19 | INDEX RANGE SCAN | AGILE_FLEX_UQ | 1 | | 1 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------ Predicate Information (identified by operation id):
--------------------------------------------------- 7 - filter("U"."EMAIL"='fanjiangen@kedacom.com')
8 - filter(("SO"."REQUIRED"=1 OR "SO"."REQUIRED"=5) AND ("SO"."SIGNOFF_STATUS"=0 OR
"SO"."SIGNOFF_STATUS"=4))
9 - access("U"."ID"="SO"."USER_ASSIGNED")
10 - filter("WP"."CHANGED_BY" IS NULL)
11 - access("WP"."ID"="SO"."PROCESS_ID")
12 - filter(("CH"."SUBCLASS"=1 OR "CH"."SUBCLASS"=2473495 OR "CH"."SUBCLASS"=2473519 OR
"CH"."SUBCLASS"=2473531 OR "CH"."SUBCLASS"=2473549 OR "CH"."SUBCLASS"=2473579 OR
"CH"."SUBCLASS"=2474897 OR "CH"."SUBCLASS"=2475794 OR "CH"."SUBCLASS"=2478248 OR
"CH"."SUBCLASS"=2479577 OR "CH"."SUBCLASS"=2479783 OR "CH"."SUBCLASS"=2480885 OR
"CH"."SUBCLASS"=2485198) AND "CH"."STATUS"="WP"."STATE")
13 - access("CH"."ID"="WP"."CHANGE_ID")
14 - filter("CH"."SUBCLASS"=1 OR "CH"."SUBCLASS"=2473495 OR "CH"."SUBCLASS"=2473519 OR
"CH"."SUBCLASS"=2473531 OR "CH"."SUBCLASS"=2473549 OR "CH"."SUBCLASS"=2473579 OR
"CH"."SUBCLASS"=2474897 OR "CH"."SUBCLASS"=2475794 OR "CH"."SUBCLASS"=2479577 OR
"CH"."SUBCLASS"=2479783 OR "CH"."SUBCLASS"=2480885 OR "CH"."SUBCLASS"=2485198 OR
"CH"."SUBCLASS"=2478248 AND "L"."VALUE"='审批')
15 - access("L"."LANGID"=4 AND "L"."TYPE"=4450 AND "CH"."STATUS"="L"."ID")
17 - access("CH"."ORIGINATOR"="U1"."ID")
19 - access("F"."ID"(+)="CH"."ID" AND "F"."ATTID"(+)=2017)
filter("F"."ATTID"(+)=2017) Statistics
-----------------------------------------------------------
1 DB time
1 Requests to/from client
2 non-idle wait count
1 opened cursors cumulative
1 opened cursors current
1 pinned cursors current
58008 session uga memory
2 user calls
从解释计划中可以看出有2个地方预估时间很长，一个是对agileuser用户表，另一个是signoff，用户审批表。
用户表总共也就几千条记录，而且邮箱还是唯一的，所以资源耗费不是很大。而signoff表数据基数相当庞大，达到百万级。
虽然已经走了索引了，但是走的索引是审批用户的ID这个字段。对于绝大部分人来说，其下的数据不是很多。所以查询起来不会很慢。
但对于某些重要领导（大部分审批流都要他批的那种）数据量就会变得相当庞大。而这个SQL执行慢也恰好是遇到了这样的领导才会慢。
经过具体问题，根据某领导张三的用户ID可以在这张表中查询到2万多数据。也就是说数据库引擎根据现有索引找到该用户的记录后，还要从2万多记录中找到当前需要审批的记录。
分析至此，考虑将用户id以及sql语句中另外的2个条件加入组合索引。重新使用autotrace分析语句，结果如下：
Elapsed: 00:00:00.008
Plan hash value: 441713506 -------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 24 | 5280 | 622 (1)| 00:00:08 |
| 1 | NESTED LOOPS OUTER | | 24 | 5280 | 622 (1)| 00:00:08 |
| 2 | NESTED LOOPS | | 22 | 3564 | 589 (1)| 00:00:08 |
| 3 | NESTED LOOPS | | 22 | 3234 | 578 (1)| 00:00:07 |
| 4 | NESTED LOOPS | | 56 | 6328 | 550 (1)| 00:00:07 |
| 5 | NESTED LOOPS | | 156 | 11076 | 472 (1)| 00:00:06 |
| 6 | NESTED LOOPS | | 212 | 10176 | 260 (1)| 00:00:04 |
|* 7 | TABLE ACCESS FULL | AGILEUSER | 1 | 29 | 168 (1)| 00:00:03 |
| 8 | INLIST ITERATOR | | | | | |
| 9 | TABLE ACCESS BY INDEX ROWID| SIGNOFF | 210 | 3990 | 93 (2)| 00:00:02 |
|* 10 | INDEX RANGE SCAN | SIGNOFF_IDX_KD_1 | 210 | | 3 (34)| 00:00:01 |
|* 11 | TABLE ACCESS BY INDEX ROWID | WORKFLOW_PROCESS | 1 | 23 | 1 (0)| 00:00:01 |
|* 12 | INDEX UNIQUE SCAN | WF_PROCESS_PK | 1 | | 1 (0)| 00:00:01 |
|* 13 | TABLE ACCESS BY INDEX ROWID | CHANGE | 1 | 42 | 1 (0)| 00:00:01 |
|* 14 | INDEX UNIQUE SCAN | CHANGE_PK | 1 | | 1 (0)| 00:00:01 |
|* 15 | TABLE ACCESS BY INDEX ROWID | LANGTABLE | 1 | 34 | 1 (0)| 00:00:01 |
|* 16 | INDEX UNIQUE SCAN | LANGTABLE_PK | 1 | | 1 (0)| 00:00:01 |
| 17 | TABLE ACCESS BY INDEX ROWID | AGILEUSER | 1 | 15 | 1 (0)| 00:00:01 |
|* 18 | INDEX UNIQUE SCAN | AGILEUSER_PK | 1 | | 1 (0)| 00:00:01 |
| 19 | TABLE ACCESS BY INDEX ROWID | AGILE_FLEX | 1 | 58 | 2 (0)| 00:00:01 |
|* 20 | INDEX RANGE SCAN | AGILE_FLEX_UQ | 1 | | 1 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id):
--------------------------------------------------- 7 - filter("U"."EMAIL"='fanjiangen@kedacom.com')
10 - access("U"."ID"="SO"."USER_ASSIGNED" AND ("SO"."SIGNOFF_STATUS"=0 OR
"SO"."SIGNOFF_STATUS"=4) AND ("SO"."REQUIRED"=1 OR "SO"."REQUIRED"=5))
11 - filter("WP"."CHANGED_BY" IS NULL)
12 - access("WP"."ID"="SO"."PROCESS_ID")
13 - filter(("CH"."SUBCLASS"=1 OR "CH"."SUBCLASS"=2473495 OR "CH"."SUBCLASS"=2473519 OR
"CH"."SUBCLASS"=2473531 OR "CH"."SUBCLASS"=2473549 OR "CH"."SUBCLASS"=2473579 OR
"CH"."SUBCLASS"=2474897 OR "CH"."SUBCLASS"=2475794 OR "CH"."SUBCLASS"=2478248 OR
"CH"."SUBCLASS"=2479577 OR "CH"."SUBCLASS"=2479783 OR "CH"."SUBCLASS"=2480885 OR
"CH"."SUBCLASS"=2485198) AND "CH"."STATUS"="WP"."STATE")
14 - access("CH"."ID"="WP"."CHANGE_ID")
15 - filter("CH"."SUBCLASS"=1 OR "CH"."SUBCLASS"=2473495 OR "CH"."SUBCLASS"=2473519 OR
"CH"."SUBCLASS"=2473531 OR "CH"."SUBCLASS"=2473549 OR "CH"."SUBCLASS"=2473579 OR
"CH"."SUBCLASS"=2474897 OR "CH"."SUBCLASS"=2475794 OR "CH"."SUBCLASS"=2479577 OR
"CH"."SUBCLASS"=2479783 OR "CH"."SUBCLASS"=2480885 OR "CH"."SUBCLASS"=2485198 OR
"CH"."SUBCLASS"=2478248 AND "L"."VALUE"='审批')
16 - access("L"."LANGID"=4 AND "L"."TYPE"=4450 AND "CH"."STATUS"="L"."ID")
18 - access("CH"."ORIGINATOR"="U1"."ID")
20 - access("F"."ID"(+)="CH"."ID" AND "F"."ATTID"(+)=2017)
filter("F"."ATTID"(+)=2017) Statistics
-----------------------------------------------------------
1 CPU used by this session
1 Requests to/from client
2 non-idle wait count
1 opened cursors cumulative
1 opened cursors current
1 pinned cursors current
2 user calls
>>Query Run In:查询结果 3
从上面的结果可以看出，sql走入了新建的组合索引（SIGNOFF_IDX_KD_1），预估执行时间从原来的06秒变成02秒，cost也大幅度下降，整句SQL的执行时间也从4秒多下降到了200ms。
由此可见，索引对于查询的优化真是显而易见。
记一次SQL性能优化,查询时间从4000ms优化到200ms.的更多相关文章
- sql exist 优化查询时间
1.非exist,查询需要20多秒 2.使用exist后 3.表连接也能优化
- PHP获取MySQL执行sql语句的查询时间
//计时开始 runtime(); //执行查询 mysql_query($sql); //计时结束. echo runtime(1); //计时函数 function runtime($mode=0 ...
- sql server获取查询时间
declare @d datetime set @d=getdate() /*你的SQL脚本开始*/ /*你的SQL脚本结束*/ select [语句执行花费时间(毫秒)]=datediff(ms,@ ...
- 计算sql语句的查询时间
set statistics profile on set statistics io on set statistics time on go <这里写上你的语句...> go set ...
- （转载）PHP怎么获取MySQL执行sql语句的查询时间
(转载自CSDN) 方法一: //计时开始 runtime(); //执行查询 mysql_query($sql); //计时结束. echo runtime(1); //计时函数 function ...
- 记一次sql server 性能调优，查询从20秒至2秒
一.需求 需求很简单,就是需要查询一个报表,只有1个表,数据量大约60万左右,但是中间有些逻辑. 先说明一下服务器配置情况:1核CPU.2GB内存.机械硬盘.Sqlserver 2008 R2.Win ...
- SQL性能优化-查询条件与字段分开执行，union代替in与or，存储过程代替union
PS:概要.背景.结语都是日常“装X”,可以跳过直接看优化历程 环境:SQL Server 2008 R2.阿里云RDS:辅助工具:SQL 审计 概要 一个订单列表分页查询功能,单从SQL性能来讲,从 ...
- SQL Server 2016 查询存储性能优化小结
SQL Server 2016已经发布了有半年多,相信还有很多小伙伴还没有开始使用,今天我们来谈谈SQL Server 2016 查询存储性能优化,希望大家能够喜欢 作为一个DBA,排除SQL Ser ...
- 性能优化-查询最耗CPU的SESSION与SQL
在linux 系统中 用top命令查出CPU最高的SPID,再将SPID给存储过程,可以查出该进程的SQLTEXT create or replace procedure pro_get_sqltex ...
随机推荐
- Centos7.4安装配置haproxy和Keepalived补充内容
补充比较杂 1.当master服务恢复正常之后,backup机器收到消息,然后让出vip 下面是master机器服务恢复正常后,backup机器的Keepalived日志 收到master的消息通知, ...
- CSS 社区的解决方案，对比
在众多解决方案中,没有绝对的优劣.还是要结合自己的场景来决定. 我们团队在使用过 scss 和 css modules 后,仍然又重新选择了使用 scss.css modules 虽然有效解决了样式冲 ...
- 汉语言处理工具pyhanlp的拼音转换与字符正则化
汉字转拼音 HanLP中的汉字转拼音功能也十分的强大. 说明: l HanLP不仅支持基础的汉字转拼音,还支持声母.韵母.音调.音标和输入法首字母首声母功能. l HanLP能够识别多音字,也能给繁体 ...
- OkHttp官方中文文档
https://blog.csdn.net/jackingzheng/article/details/51778793 https://www.cnblogs.com/ldq2016/p/879630 ...
- debian下arp欺骗
sudo sysctl -w net.ipv4.ip_forward= sudo sysctl -p arpspoof -i eth0 -t 目标ip -r 伪装ip或者ettercap -i eth ...
- Python程序打包之PyInstaller
1.背景说明 [Python版本]Python 2.7.14 [系统平台]Windows 7 [优缺点描述]据说PyInstaller 比较慢,但是PyInstaller打包出来的exe简洁(就一个文 ...
- os.rename 和os.replace
f1 = open("hello.txt","w") f1.write("hello,my name is bobo.") f1.close ...
- go api json 输出尝试
package main import ( "fmt" "net/http" //"net/url" "github.com/dr ...
- liunx 部署 spring boot
Xshell for Xmanager Enterprise 4 (Build 0232)Copyright (c) 2002-2014 NetSarang Computer, Inc. All ri ...
- 【JVM底层策略 一】GC roots如何判断对象不可达
查找内存中不再使用的对象 引用计数法 引用计数法就是如果一个对象没有被任何引用指向,则可视之为垃圾.这种方法的缺点就是不能检测到环的存在. 2.根搜索算法 根搜索算法的基本思路就是通过一系列名为”GC ...