一次性能优化将filter转换
有一条SQL性能有问题，在运行计划中发现filter。遇到它要小心了，类似于nestloop。我曾经的blog对它有研究探索运行计划中filter的原理.用exists极易引起filter.
优化前：
SELECT GGPI.ID,
GGPI.PLAN_STATE,
GGPI.PLAN_TYPE,
GGPI.PLAN_CODE,
GGPI.PLAN_SOURCE_TYPE,
GGPI.PLAN_BEGIN_DATE,
GGPI.PLAN_END_DATE,
GGPI.REAL_BEGIN_DATE,
GGPI.REAL_END_DATE,
GGPI.WORK_MASTER_UNAME,
GGPI.WORK_MASTER_UID,
GGPI.WORK_TEAM_ONAME,
GGPI.WORK_SITE_NAMES,
GGPI.WORK_CONTENT,
GGPI.WORK_TYPE,
GGPI.WORK_MEMBER_UNAMES,
GGPI.DIGGATCH_WORK_UNAME,
GGPI.WORKING
FROM GG_PD_PP_INFO GGPI
WHERE PLAN_STATE IN (50, 60, 70)
AND ((GGPI.PLAN_BEGIN_DATE >= sysdate and GGPI.PLAN_BEGIN_DATE <= sysdate) or
(GGPI.PLAN_END_DATE >= sysdate and GGPI.PLAN_END_DATE <= sysdate) or
(GGPI.PLAN_BEGIN_DATE < sysdate and GGPI.PLAN_END_DATE > sysdate))
AND GGPI.BUREAU_CODE = '0306'
AND (GGPI.WORK_MASTER_UID ='FA3502D1291A4A07A3B9E3E0F9A41904' OR EXISTS
(SELECT 1
FROM GG_PD_PP_WORK_MEMBER PMEM
WHERE PMEM.PROD_PLAN_ID = GGPI.ID
AND PMEM.WORK_MEMBER_UID = 'FA3502D1291A4A07A3B9E3E0F9A41904'
AND PMEM.BUREAU_CODE = '0306'));
Execution Plan
----------------------------------------------------------
Plan hash value: 980323934
-------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
-------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9670 | 2039K| 25694 (1)| 00:05:09 | | |
|* 1 | FILTER | | | | | | | |
| 2 | PARTITION LIST SINGLE | | 72882 | 15M| 25694 (1)| 00:05:09 | KEY | KEY |
| 3 | PARTITION RANGE ALL | | 72882 | 15M| 25694 (1)| 00:05:09 | 1 | 14 |
|* 4 | TABLE ACCESS FULL | GG_PD_PP_INFO | 72882 | 15M| 25694 (1)| 00:05:09 | | |
|* 5 | TABLE ACCESS BY GLOBAL INDEX ROWID| GG_PD_PP_WORK_MEMBER | 1 | 49 | 5 (0)| 00:00:01 | 4 | 4 |
|* 6 | INDEX RANGE SCAN | IDX_PP_WORK_MEMBER_PPID | 6 | | 3 (0)| 00:00:01 | | |
-------------------------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("GGPI"."WORK_MASTER_UID" IS NULL OR EXISTS (SELECT 0 FROM "GG_PD_PP_WORK_MEMBER"
"PMEM" WHERE "PMEM"."PROD_PLAN_ID"=:B1 AND "PMEM"."WORK_MEMBER_UID"='FA3502D1291A4A07A3B9E3E0F9A41904' AND
"PMEM"."BUREAU_CODE"='0306'))
4 - filter(("PLAN_STATE"=50 OR "PLAN_STATE"=60 OR "PLAN_STATE"=70) AND ("GGPI"."PLAN_END_DATE">SYSDATE@! AND
"GGPI"."PLAN_BEGIN_DATE"<SYSDATE@! OR "GGPI"."PLAN_BEGIN_DATE"=SYSDATE@! OR "GGPI"."PLAN_END_DATE"=SYSDATE@!))
5 - filter("PMEM"."WORK_MEMBER_UID"='FA3502D1291A4A07A3B9E3E0F9A41904' AND "PMEM"."BUREAU_CODE"='0306')
6 - access("PMEM"."PROD_PLAN_ID"=:B1)
Statistics
----------------------------------------------------------
1 recursive calls
0 db block gets
103763 consistent gets
0 physical reads
0 redo size
3178 bytes sent via SQL*Net to client
524 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
7 rows processed
优化后：
SELECT GGPI.ID,
GGPI.PLAN_STATE,
GGPI.PLAN_TYPE,
GGPI.PLAN_CODE,
GGPI.PLAN_SOURCE_TYPE,
GGPI.PLAN_BEGIN_DATE,
GGPI.PLAN_END_DATE,
GGPI.REAL_BEGIN_DATE,
GGPI.REAL_END_DATE,
GGPI.WORK_MASTER_UNAME,
GGPI.WORK_MASTER_UID,
GGPI.WORK_TEAM_ONAME,
GGPI.WORK_SITE_NAMES,
GGPI.WORK_CONTENT,
GGPI.WORK_TYPE,
GGPI.WORK_MEMBER_UNAMES,
GGPI.DIGGATCH_WORK_UNAME,
GGPI.WORKING
FROM GG_PD_PP_INFO GGPI
WHERE PLAN_STATE IN (50, 60, 70)
AND ((GGPI.PLAN_BEGIN_DATE >= sysdate and GGPI.PLAN_BEGIN_DATE <= sysdate) or
(GGPI.PLAN_END_DATE >= sysdate and GGPI.PLAN_END_DATE <= sysdate) or
(GGPI.PLAN_BEGIN_DATE < sysdate and GGPI.PLAN_END_DATE > sysdate))
AND GGPI.BUREAU_CODE = '0306'
AND GGPI.WORK_MASTER_UID ='FA3502D1291A4A07A3B9E3E0F9A41904'
union all
SELECT GGPI.ID,
GGPI.PLAN_STATE,
GGPI.PLAN_TYPE,
GGPI.PLAN_CODE,
GGPI.PLAN_SOURCE_TYPE,
GGPI.PLAN_BEGIN_DATE,
GGPI.PLAN_END_DATE,
GGPI.REAL_BEGIN_DATE,
GGPI.REAL_END_DATE,
GGPI.WORK_MASTER_UNAME,
GGPI.WORK_MASTER_UID,
GGPI.WORK_TEAM_ONAME,
GGPI.WORK_SITE_NAMES,
GGPI.WORK_CONTENT,
GGPI.WORK_TYPE,
GGPI.WORK_MEMBER_UNAMES,
GGPI.DIGGATCH_WORK_UNAME,
GGPI.WORKING
FROM GG_PD_PP_INFO GGPI
WHERE PLAN_STATE IN (50, 60, 70)
AND ((GGPI.PLAN_BEGIN_DATE >= sysdate and GGPI.PLAN_BEGIN_DATE <= sysdate) or
(GGPI.PLAN_END_DATE >= sysdate and GGPI.PLAN_END_DATE <= sysdate) or
(GGPI.PLAN_BEGIN_DATE < sysdate and GGPI.PLAN_END_DATE > sysdate))
AND GGPI.BUREAU_CODE = '0306'
AND GGPI.ID in
(SELECT PMEM.PROD_PLAN_ID
FROM GG_PD_PP_WORK_MEMBER PMEM
WHERE PMEM.WORK_MEMBER_UID = 'FA3502D1291A4A07A3B9E3E0F9A41904'
AND PMEM.BUREAU_CODE = '0306');
Execution Plan
----------------------------------------------------------
Plan hash value: 1911592856
-----------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
-----------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1227 | 316K| 3740 (1)| 00:00:45 | | |
| 1 | UNION-ALL | | | | | | | |
|* 2 | TABLE ACCESS BY GLOBAL INDEX ROWID | GG_PD_PP_INFO | 22 | 4752 | 160 (0)| 00:00:02 | ROWID | ROWID |
|* 3 | INDEX RANGE SCAN | IND_GGPI_WORK_MASTER_UID | 222 | | 5 (0)| 00:00:01 | | |
| 4 | NESTED LOOPS | | 1205 | 311K| 3580 (1)| 00:00:43 | | |
| 5 | NESTED LOOPS | | 1205 | 311K| 3580 (1)| 00:00:43 | | |
| 6 | SORT UNIQUE | | 1205 | 59045 | 2373 (1)| 00:00:29 | | |
|* 7 | TABLE ACCESS BY GLOBAL INDEX ROWID| GG_PD_PP_WORK_MEMBER | 1205 | 59045 | 2373 (1)| 00:00:29 | 4 | 4 |
|* 8 | INDEX RANGE SCAN | IND_GGPW_WORK_MEMBER_UID | 4240 | | 31 (0)| 00:00:01 | | |
|* 9 | INDEX UNIQUE SCAN | PK_GG_PD_PP_INFO | 1 | | 1 (0)| 00:00:01 | | |
|* 10 | TABLE ACCESS BY GLOBAL INDEX ROWID | GG_PD_PP_INFO | 1 | 216 | 2 (0)| 00:00:01 | ROWID | ROWID |
-----------------------------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - filter(("PLAN_STATE"=50 OR "PLAN_STATE"=60 OR "PLAN_STATE"=70) AND ("GGPI"."PLAN_END_DATE">SYSDATE@! AND
"GGPI"."PLAN_BEGIN_DATE"<SYSDATE@! OR "GGPI"."PLAN_BEGIN_DATE"=SYSDATE@! OR "GGPI"."PLAN_END_DATE"=SYSDATE@!) AND
"GGPI"."BUREAU_CODE"='0306')
3 - access("GGPI"."WORK_MASTER_UID"='FA3502D1291A4A07A3B9E3E0F9A41904')
7 - filter("PMEM"."BUREAU_CODE"='0306')
8 - access("PMEM"."WORK_MEMBER_UID"='FA3502D1291A4A07A3B9E3E0F9A41904')
9 - access("GGPI"."ID"="PMEM"."PROD_PLAN_ID")
10 - filter(("GGPI"."PLAN_END_DATE">SYSDATE@! AND "GGPI"."PLAN_BEGIN_DATE"<SYSDATE@! OR
"GGPI"."PLAN_BEGIN_DATE"=SYSDATE@! OR "GGPI"."PLAN_END_DATE"=SYSDATE@!) AND ("PLAN_STATE"=50 OR "PLAN_STATE"=60 OR
"PLAN_STATE"=70) AND "GGPI"."BUREAU_CODE"='0306')
Statistics
----------------------------------------------------------
1 recursive calls
0 db block gets
4321 consistent gets
550 physical reads
0 redo size
3228 bytes sent via SQL*Net to client
524 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
7 rows processed
一次性能优化将filter转换的更多相关文章
- 一次性能优化，tps从400+到4k+
项目介绍 路由网关项目watchman ,接收前端http请求转发到后端业务系统,功能安全验证,限流,转发. 使用技术:spring boot+ nreflix zuul,最开始日志使用slf4j+l ...
- MYSQL一次性能优化实战经历[转]
每次经历数据库性能调优,都是对性能优化的再次认识.对自己知识不足的有力验证,只有不断总结.学习才能少走弯路. 一.性能问题描述 应用端反应系统查询缓慢,长时间出不来结果.SQLServer数据库服务器 ...
- 记一次性能优化，限制tcp_timewait数量，快速回收和重用
前言 这篇文章的主题是记录一次Python程序的性能优化,在优化的过程中遇到的问题,以及如何去解决的.为大家提供一个优化的思路,首先要声明的一点是,我的方式不是唯一的,大家在性能优化之路上遇到的问题都 ...
- 排除maven jar冲突 maven tomcat插件启动报错 filter转换异常
最近在搞一个ssm+shiro的整合 用的maven tomcat插件 启动的时候报错,提示 maven org.springframework.web.filter.CharacterEncodin ...
- 也记一次性能优化：LINQ to SQL中Contains方法的优化
距离上一篇博文更新已经两个月过去了.在此,先表一表这两个月干了些啥: 世界那么大,我也想去看看.四月份的时候,我入职了上海的一家电商公司,职位是.NET高级开发工程师.工作一个月,最大的感受是比以前小 ...
- 记一次性能优化的心酸历程【Flask+Gunicorn+pytorch+多进程+线程池，一顿操作猛如虎】
您好,我是码农飞哥,感谢您阅读本文,欢迎一键三连哦. 本文只是记录我优化的心酸历程.无他,唯记录尔.....小伙伴们可围观,可打call,可以私信与我交流. 干货满满,建议收藏,需要用到时常看看. 小 ...
- [编织消息框架][设计协议]优化long,int转换
理论部分 一个long占8byte,大多数应用业数值不超过int每次传输多4byte会很浪费 有没有什么办法可以压缩long或int呢? 答案是有的,原理好简单,如果数值不超过int.max_valu ...
- Lazy<T>在Entity Framework中的性能优化实践(附源码)
在使用EF的过程中,导航属性的lazy load机制,能够减少对数据库的不必要的访问.只有当你使用到导航属性的时候,才会访问数据库.但是这个只是对于单个实体而言,而不适用于显示列表数据的情况. 这篇文 ...
- Lazy<T>在Entity Framework中的性能优化实践
Lazy<T>在Entity Framework中的性能优化实践(附源码) 2013-10-27 18:12 by JustRun, 328 阅读, 4 评论, 收藏, 编辑 在使用EF的 ...
随机推荐
- 使用Java语言开发微信公众平台(三)
在上一节课程中,我们来学习了微信公众平台最基础的一个接口——access_token,并且能够从微信公众平台中取到access_token. 那么,在本节课程中,我们要以上节课获取到的 ...
- chattr---文件隐藏属性
- marquee
marquee: 页面的自动滚动效果,可由javascript来实现,但是今天无意中发现了一个html标签 - <marquee></marquee>可以实现多种滚动效果,无需 ...
- 【Hello 2018 D】Too Easy Problems
[链接] 我是链接,点我呀:) [题意] 在这里输入题意 [题解] 可以考虑把所有的题目按照ai排序. 然后顺序考虑最后做出来的题目个数和第i道题目的ai一样. 则1..i-1这些题目就没有用了. 值 ...
- 【2017 Multi-University Training Contest - Team 5】Rikka with Subset
[Link]: [Description] 给你a数组的n个数的所有2^n个子集的2^n个子集元素的和; 子集元素的和最大为m; 告诉你各个子集元素的和出现的次数; 如 1 2 则0出现1次,1出现1 ...
- hdu 1588 Gauss Fibonacci（矩阵嵌矩阵）
题目大意: 求出斐波那契中的 第 k*i+b 项的和. 思路分析: 定义斐波那契数列的矩阵 f(n)为斐波那契第n项 F(n) = f(n+1) f(n) 那么能够知道矩阵 A = 1 1 1 0 ...
- inflater-布局转化实现
通俗的说,inflate就相当于将一个xml中定义的布局找出来.因为在一个Activity里如果直接用findViewById()的话,对应的是setConentView()的那个layout里的组件 ...
- arm-linux-gcc 命令未找到问题
解决方法: 1.先打开一个超级用户权限的shell: 命令: ubuntu :sudo –s centos :su - 2.在当前shell下,设置环境变量: 命令:gedit /etc/profil ...
- WPF 入门《常用控件》
1.GroupBox 注意: GroupBox仍然需要布局容器来放置元素.如: StackPanel面板 <GroupBox Header="select number?"& ...
- 洛谷 P1760 通天之汉诺塔
P1760 通天之汉诺塔 题目背景 直达通天路·小A历险记第四篇 题目描述 在你的帮助下,小A成功收集到了宝贵的数据,他终于来到了传说中连接通天路的通天山.但是这距离通天路仍然有一段距离,但是小A突然 ...