SQL优化----如何使用工具快速诊断出统计信息有问题？

最新推荐文章于 2024-04-15 18:59:54 发布

深圳gg

最新推荐文章于 2024-04-15 18:59:54 发布

阅读量1.3k

点赞数

分类专栏： Oracle SQL优化

本文链接：https://blog.csdn.net/stevendbaguo/article/details/17122811

版权

Oracle SQL优化专栏收录该内容

130 篇文章 7 订阅

订阅专栏

下列的SQL语句是一个报表统计的SQL，听开发说with里面返回的记录其实很少，只有十几条而已。

SQL> WITH STAT AS (

2 SELECT PTI.FROM_BUREAU_NO,PTI.TASK_ID, PPE.PATROL_EQUIPMENT_ID,PTI.IS_COMPLETE,PPE.DEVICE_ID
3 FROM GG_pTASK_PATROL_DETAIL PTPD, GG_PTASK_ITEM PTI,GG_PATROL_EQUIPMENT PPE
4 WHERE PTPD.PATROL_LINE_ID=PPE.PATROL_LINE_ID AND PTPD.PATROL_LINE_ID IS NOT NULL
5 AND PTPD.PATROL_SUB_TYPE=1
6 AND PTPD.PATROL_TYPE=1
7 AND PTI.TASK_ID=PTPD.TASK_ID
8 AND PTI.TASK_KIND=1
9 )
10 SELECT FROM_BUREAU_NO,VOLTAGE_LEVEL,COUNT(*) allrows, SUM(REC) FROM(
11 SELECT T.FROM_BUREAU_NO,PD.DEVICE_ID,PD.VOLTAGE_LEVEL,DECODE(T.IS_COMPLETE,3,1,0) REC FROM GG_DEVICE PD,STAT T WHERE PD.DEVICE_ID=T.DEVICE_ID)
12 GROUP BY FROM_BUREAU_NO,VOLTAGE_LEVEL;
已用时间: 00: 01: 13.24 --非常慢
执行计划
----------------------------------------------------------
Plan hash value: 1646606100
--------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
--------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 253 | 15939 | | 73290 (1)| 00:14:40 |
| 1 | HASH GROUP BY | | 253 | 15939 | | 73290 (1)| 00:14:40 |
|* 2 | HASH JOIN | | 107K| 6614K| 6592K| 73283 (1)| 00:14:40 |
|* 3 | HASH JOIN | | 105K| 5354K| | 1625 (1)| 00:00:20 |
| 4 | NESTED LOOPS | | 375 | 13125 | | 806 (1)| 00:00:10 |
|* 5 | TABLE ACCESS FULL | GG_PTASK_PATROL_DETAIL | 375 | 7500 | | 55 (0)| 00:00:01 |
|* 6 | TABLE ACCESS BY INDEX ROWID| GG_PTASK_ITEM | 1 | 15 | | 2 (0)| 00:00:01 |
|* 7 | INDEX UNIQUE SCAN | PK_GG_PTASK_ITEM | 1 | | | 1 (0)| 00:00:01 |
| 8 | TABLE ACCESS FULL | GG_PATROL_EQUIPMENT | 192K| 3191K| | 818 (1)| 00:00:10 |
| 9 | TABLE ACCESS FULL | GG_DEVICE | 5226K| 54M| | 65586 (1)| 00:13:08 |
--------------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("PD"."DEVICE_ID"=TO_NUMBER("PPE"."DEVICE_ID"))
3 - access("PTPD"."PATROL_LINE_ID"="PPE"."PATROL_LINE_ID")
5 - filter("PTPD"."PATROL_SUB_TYPE"=1 AND "PTPD"."PATROL_TYPE"=1 AND "PTPD"."PATROL_LINE_ID" IS NOT NULL)
6 - filter("PTI"."TASK_KIND"=1)
7 - access("PTI"."TASK_ID"="PTPD"."TASK_ID")
统计信息
----------------------------------------------------------
1 recursive calls
0 db block gets
302451 consistent gets
176347 physical reads
0 redo size
543 bytes sent via SQL*Net to client
350 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

--开始使用利器
SQL> alter session set statistics_level=all;
SQL> set pagesize 100
SQL> SELECT FROM_BUREAU_NO,VOLTAGE_LEVEL,COUNT(*) allrows, SUM(REC) FROM(
2 SELECT PTI.FROM_BUREAU_NO,PTI.TASK_ID, PPE.PATROL_EQUIPMENT_ID,
3 DECODE(PTI.IS_COMPLETE,3,1,0) REC,PPE.DEVICE_ID,PD.VOLTAGE_LEVEL
4 FROM GG_pTASK_PATROL_DETAIL PTPD, GG_PTASK_ITEM PTI,GG_PATROL_EQUIPMENT PPE,GG_DEVICE PD
5 WHERE PTPD.PATROL_LINE_ID=PPE.PATROL_LINE_ID AND PTPD.PATROL_LINE_ID IS NOT NULL --关联路线
6 AND PTPD.PATROL_SUB_TYPE=1
7 AND PTPD.PATROL_TYPE=1
8 AND PTI.TASK_ID=PTPD.TASK_ID
9 AND PTI.TASK_KIND=1
10 AND PD.DEVICE_ID= PPE.DEVICE_ID)
11 GROUP BY FROM_BUREAU_NO,VOLTAGE_LEVEL;
FROM_BUREAU_NO VOLTAGE_LEVEL ALLROWS SUM(REC)
-------------- ------------- ---------- ----------
13 110000 11 0
13 220000 3 0

已用时间: 00: 00: 27.99
SQL> select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));
PLAN_TABLE_OUTPUT
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID ghhnk0gkny184, child number 1
-------------------------------------
SELECT FROM_BUREAU_NO,VOLTAGE_LEVEL,COUNT(*) allrows, SUM(REC) FROM( SELECT PTI.FROM_BUREAU_NO,PTI.TASK_ID, PPE.PATROL_EQUIPMENT_ID,
DECODE(PTI.IS_COMPLETE,3,1,0) REC,PPE.DEVICE_ID,PD.VOLTAGE_LEVEL FROM GG_pTASK_PATROL_DETAIL PTPD, GG_PTASK_ITEM
PTI,GG_PATROL_EQUIPMENT PPE,GG_DEVICE PD WHERE PTPD.PATROL_LINE_ID=PPE.PATROL_LINE_ID AND PTPD.PATROL_LINE_ID IS NOT NULL AND
PTPD.PATROL_SUB_TYPE=1 AND PTPD.PATROL_TYPE=1 AND PTI.TASK_ID=PTPD.TASK_ID AND PTI.TASK_KIND=1 AND
PD.DEVICE_ID= PPE.DEVICE_ID) GROUP BY FROM_BUREAU_NO,VOLTAGE_LEVEL

Plan hash value: 1646606100

--E-Rows是oracle评估的条数，A-Rows是实际返回的条数，这就是统计信息有问题的证据。

----------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts| E-Rows | A-Rows | A-Time | Buffers | Reads | OMem | 1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------------------------------------------------
| 1 | HASH GROUP BY | | 1| 253 | 2|00:00:27.97 | 302K| 78286 | | | |
|* 2 | HASH JOIN | | 1| 104K| 14|00:00:27.97 | 302K| 78286 | 1114K| 1114K| 1616K (0)|
|* 3 | HASH JOIN | | 1| 102K| 17|00:00:00.34 | 4314 | 0 | 1114K| 1114K| 441K (0)|
| 4 | NESTED LOOPS | | 1| 375| 2|00:00:00.01 | 409 | 0 | | | |
|* 5 | TABLE ACCESS FULL | GG_PTASK_PATROL_DETAIL | 1| 375| 55|00:00:00.01 | 242 | 0 | | | |
|* 6 | TABLE ACCESS BY INDEX ROWID| GG_PTASK_ITEM | 55| 1 | 2 |00:00:00.01 | 167 | 0 | | | |
|* 7 | INDEX UNIQUE SCAN | PK_GG_PTASK_ITEM | 55| 1 | 55 |00:00:00.01 | 112 | 0 | | | |
| 8 | TABLE ACCESS FULL | GG_PATROL_EQUIPMENT | 1| 190K| 190K|00:00:00.01 | 3905 | 0 | | | |
| 9 | TABLE ACCESS FULL | GG_DEVICE | 1| 5226K| 5226K|00:00:20.91 | 298K| 78286 | | | |
----------------------------------------------------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
--------------------------------------------------
2 - access("PD"."DEVICE_ID"="PPE"."DEVICE_ID")
3 - access("PTPD"."PATROL_LINE_ID"="PPE"."PATROL_LINE_ID")
5 - filter(("PTPD"."PATROL_SUB_TYPE"=1 AND "PTPD"."PATROL_TYPE"=1 AND "PTPD"."PATROL_LINE_ID" IS NOT NULL))
6 - filter("PTI"."TASK_KIND"=1)
7 - access("PTI"."TASK_ID"="PTPD"."TASK_ID")
已选择33行。

SQL> EXEC dbms_stats.gather_table_stats(USER,'GG_PTASK_ITEM',CASCADE=>TRUE);
SQL> EXEC dbms_stats.gather_table_stats(USER,'GG_PATROL_EQUIPMENT',CASCADE=>TRUE);
SQL> EXEC dbms_stats.gather_table_stats(USER,'GG_DEVICE',CASCADE=>TRUE);
SQL> EXEC dbms_stats.gather_table_stats(USER,'GG_pTASK_PATROL_DETAIL',CASCADE=>TRUE);

SQL> WITH STAT AS (
2 SELECT PTI.FROM_BUREAU_NO,PTI.TASK_ID, PPE.PATROL_EQUIPMENT_ID,PTI.IS_COMPLETE,PPE.DEVICE_ID
3 FROM GG_pTASK_PATROL_DETAIL PTPD, GG_PTASK_ITEM PTI,GG_PATROL_EQUIPMENT PPE
4 WHERE PTPD.PATROL_LINE_ID=PPE.PATROL_LINE_ID AND PTPD.PATROL_LINE_ID IS NOT NULL
5 AND PTPD.PATROL_SUB_TYPE=1
6 AND PTPD.PATROL_TYPE=1
7 AND PTI.TASK_ID=PTPD.TASK_ID
8 AND PTI.TASK_KIND=1
9 )
10 SELECT FROM_BUREAU_NO,VOLTAGE_LEVEL,COUNT(*) allrows, SUM(REC) FROM(
11 SELECT T.FROM_BUREAU_NO,PD.DEVICE_ID,PD.VOLTAGE_LEVEL,DECODE(T.IS_COMPLETE,3,1,0) REC
12 FROM GG_DEVICE PD,STAT T WHERE PD.DEVICE_ID=T.DEVICE_ID)
13 GROUP BY FROM_BUREAU_NO,VOLTAGE_LEVEL;
已用时间: 00: 00: 00.15 --收集统计信息后，性能提高了几十倍。
执行计划
----------------------------------------------------------
Plan hash value: 3921132085
------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 238 | 14042 | 21044 (1)| 00:04:13 |
| 1 | HASH GROUP BY | | 238 | 14042 | 21044 (1)| 00:04:13 |
| 2 | NESTED LOOPS | | 10132 | 583K| 21042 (1)| 00:04:13 |
|* 3 | HASH JOIN | | 10020 | 469K| 988 (1)| 00:00:12 |
| 4 | NESTED LOOPS | | 36 | 1260 | 127 (0)| 00:00:02 |
|* 5 | TABLE ACCESS FULL | GG_PTASK_PATROL_DETAIL | 36 | 720 | 55 (0)| 00:00:01 |
|* 6 | TABLE ACCESS BY INDEX ROWID| GG_PTASK_ITEM | 1 | 15 | 2 (0)| 00:00:01 |
|* 7 | INDEX UNIQUE SCAN | PK_GG_PTASK_ITEM | 1 | | 1 (0)| 00:00:01 |
| 8 | TABLE ACCESS FULL | GG_PATROL_EQUIPMENT | 192K| 2439K| 859 (1)| 00:00:11 |
| 9 | TABLE ACCESS BY INDEX ROWID | GG_DEVICE | 1 | 11 | 2 (0)| 00:00:01 |
|* 10 | INDEX UNIQUE SCAN | PK_GG_DEVICE | 1 | | 1 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - access("PTPD"."PATROL_LINE_ID"="PPE"."PATROL_LINE_ID")
5 - filter("PTPD"."PATROL_SUB_TYPE"=1 AND "PTPD"."PATROL_TYPE"=1 AND "PTPD"."PATROL_LINE_ID" IS
NOT NULL)
6 - filter("PTI"."TASK_KIND"=1)
7 - access("PTI"."TASK_ID"="PTPD"."TASK_ID")
10 - access("PD"."DEVICE_ID"="PPE"."DEVICE_ID")
统计信息
----------------------------------------------------------
1 recursive calls
0 db block gets
4364 consistent gets
1 physical reads
0 redo size
543 bytes sent via SQL*Net to client
350 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
2 rows processed

深圳gg

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
SQL优化----如何使用工具快速诊断出统计信息有问题？

下列的SQL语句是一个报表统计的SQL，听开发说with里面返回的记录其实很少，只有十几条而已。SQL> WITH STAT AS ( 2 SELECT PTI.FROM_BUREAU_NO,PTI.TASK_ID, PPE.PATROL_EQUIPMENT_ID,PTI.IS_COMPLETE,PPE.DEVICE_ID 3 FROM GG_pTASK_PATRO
复制链接

扫一扫

专栏目录