查询统计占用了查询时间的99% [英] Query statistic is taking 99% of the query time
问题描述
我有一个SQL查询,执行该查询大约需要1.81s秒.我在phpMyAdmin中进行了分析,发现99%的时间都花在了统计"上.
I have an SQL query which take about 1.81s seconds to execute. I profiled it in phpMyAdmin and I saw that 99% of the time is spend on 'statistics'.
有趣的是,如果我再添加一个联接,则时间最多约为23秒,而如果我删除1个联接,则时间约为0.26秒.
Interestingly if I add one more join, the time goes up to about 23 seconds and if I remove 1 join the time goes down to about 0.26 seconds.
我试图将product_id =
放在联接中attribute_id =
的位置.它不会提高性能,并且统计信息"仍会在99%的时间内使用.
I have tried to put the product_id =
in the where the the attribute_id =
in the join. It does not improve the performance and the 'statistics' still use 99% of the time.
我尝试过group by p.id
,虽然确实可以提高性能,但是统计"仍然有98%的时间使用.
I have tried group by p.id
which does improve performance a little but 'statistics' still use 98% of the time.
是否有办法提高统计"性能或仅将其禁用?
注意:我在所有PK/FK上都有索引.
Note: I do have indexes on all PK/FK.
有用的信息
sqlfiddle 此处
sqlfiddle here
查询返回约150行.
ss_product table has 1k rows
ss_product_attributes has 30 rows
ss_product_attribute_varchars has 5000 rows
ss_product_attribute_decimals has 10000 rows
架构
ss_products has many ss_product_attribute_varchars
ss_products has many ss_product_attribute_decimals
ss_product_attribute_varchars belongs to ss_product_attributes
ss_product_attribute_decimals belongs to ss_product_attributes
-------------
-ss_products-
-------------
- PK - id -
-------------
--------------------------------
-ss_product_attribute_varchars-
--------------------------------
- PK - id -
- FK - product_id -
- FK - attribute_id -
--------------------------------
--------------------------------
-ss_products_attribute_decimals-
--------------------------------
- PK - id -
- FK - product_id -
- FK - attribute_id -
--------------------------------
------------------------
-ss_products_attributes-
------------------------
- PK - id -
------------------------
配置文件
Starting 101 µs
Checking Permissions 6 µs
Checking Permissions 2 µs
Checking Permissions 2 µs
Checking Permissions 2 µs
Checking Permissions 1 µs
Checking Permissions 2 µs
Checking Permissions 2 µs
Checking Permissions 2 µs
Checking Permissions 2 µs
Checking Permissions 3 µs
Opening Tables 47 µs
System Lock 10 µs
Init 35 µs
Optimizing 21 µs
Statistics 1.8 s //1.8seconds!!!!
Preparing 46 µs
Executing 3 µs
Sending Data 41 ms
End 7 µs
Query End 4 µs
Closing Tables 12 µs
Freeing Items 463 µs
Logging Slow Query 3 µs
Cleaning Up 9 µs
查询
select
aa.value as attribute_a,
bb.value as attribute_b,
cc.value as attribute_c,
dd.value as attribute_d,
ee.value as attribute_e,
ff.value as attribute_f,
gg.value as attribute_g,
hh.value as attribute_h,
ii.value as attribute_i
from ss_products as p
inner join ss_product_attribute_varchars as aa
on p.id = aa.product_id
inner join ss_product_attribute_varchars as bb
on p.id = bb.product_id
inner join ss_product_attribute_varchars as cc
on p.id = cc.product_id
inner join ss_product_attribute_decimals as dd
on p.id = dd.product_id
inner join ss_product_attribute_decimals as ee
on p.id = ee.product_id
inner join ss_product_attribute_decimals as ff
on p.id = ff.product_id
inner join ss_product_attribute_varchars as gg
on p.id = gg.product_id
INNER JOIN ss_product_attribute_varchars AS hh
ON p.id = hh.product_id
INNER JOIN ss_product_attribute_varchars AS ii
ON p.id = ii.product_id
where
aa.attribute_id = 8 AND
bb.attribute_id = 6 AND
cc.attribute_id = 7 AND
dd.attribute_id = 9 and
ee.attribute_id = 10 and
ff.attribute_id = 11 AND
gg.attribute_id = 20 AND
hh.attribute_id = 2 AND
ii.attribute_id = 3
说明
id| select_type| table | type | possible_keys |key |key_len| ref | rows| Extra
1 | SIMPLE | gg | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | attribute_id | 4 | const | 143 |
1 | SIMPLE | aa | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | fk_product_attribute_varchars_products1_idx | 4 | my_id.gg.product_id | 2 | Using where
1 | SIMPLE | cc | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | fk_product_attribute_varchars_products1_idx | 4 | my_id.gg.product_id | 2 | Using where
1 | SIMPLE | bb | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | fk_product_attribute_varchars_products1_idx | 4 | my_id.gg.product_id | 2 | Using where
1 | SIMPLE | ii | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | fk_product_attribute_varchars_products1_idx | 4 | my_id.aa.product_id | 2 | Using where
1 | SIMPLE | p | eq_ref | PRIMARY | PRIMARY | 4 | my_id.aa.product_id | 1 | Using where; Using index
1 | SIMPLE | dd | ref | fk_product_attribute_decimals_products1_idx,fk_pro... | fk_product_attribute_decimals_products1_idx | 4 | my_id.cc.product_id | 2 | Using where
1 | SIMPLE | ee | ref | fk_product_attribute_decimals_products1_idx,fk_pro... | fk_product_attribute_decimals_products1_idx | 4 | my_id.ii.product_id | 2 | Using where
1 | SIMPLE | ff | ref | fk_product_attribute_decimals_products1_idx,fk_pro... | fk_product_attribute_decimals_products1_idx | 4 | my_id.p.id | 2 | Using where
1 | SIMPLE | hh | ref | fk_product_attribute_varchars_products1_idx,fk_pro... | fk_product_attribute_varchars_products1_idx | 4 | my_id.cc.product_id | 2 | Using where
完整的个人资料(由Raymond Nijland提出)
Complete profile (as asked by Raymond Nijland)
+----------------------+----------+----------+------------+-------------------+---------------------+--------------+---------------+---------------+-------------------+-------------------+-------------------+-------+-----------------+---------------+-------------+
| Status | Duration | CPU_user | CPU_system | Context_voluntary | Context_involuntary | Block_ops_in | Block_ops_out | Messages_sent | Messages_received | Page_faults_major | Page_faults_minor | Swaps | Source_function | Source_file | Source_line |
+----------------------+----------+----------+------------+-------------------+---------------------+--------------+---------------+---------------+-------------------+-------------------+-------------------+-------+-----------------+---------------+-------------+
| starting | 0.000151 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL |
| checking permissions | 0.000005 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000002 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000003 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000002 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000003 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000002 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000005 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000002 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000002 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| checking permissions | 0.000005 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4751 |
| Opening tables | 0.000051 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_base.cc | 4838 |
| System lock | 0.000016 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | lock.cc | 299 |
| init | 0.000046 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 2560 |
| optimizing | 0.000031 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 869 |
| statistics | 1.419911 | 1.404009 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 1060 |
| preparing | 0.000050 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 1082 |
| executing | 0.000003 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 1829 |
| Sending data | 0.018508 | 0.015600 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 2371 |
| end | 0.000007 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_select.cc | 2596 |
| query end | 0.000004 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4440 |
| closing tables | 0.000018 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 4492 |
| freeing items | 0.000110 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 5640 |
| logging slow query | 0.000004 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 1461 |
| cleaning up | 0.000008 | 0.000000 | 0.000000 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | <unknown> | sql_parse.cc | 1417 |
+----------------------+----------+----------+------------+-------------------+---------------------+--------------+---------------+---------------+------------------+-------------------+-------------------+-------+-----------------+---------------+-------------+
推荐答案
提出一个答案,因为注释确实支持代码格式.
Made an answer out off it because an comment doenst support code formatting.
配置文件输出的统计信息在MySQL的C ++源代码中的此部分中定义
Statistics in profile output is defined within this part in the C++ source code off MySQL
/* Calculate how to do the join */
thd_proc_info(thd, "statistics");
if (make_join_statistics(this, select_lex->leaf_tables, conds, &keyuse) ||
thd->is_fatal_error)
{
DBUG_PRINT("error",("Error: make_join_statistics() failed"));
DBUG_RETURN(1);
}
但是您仍然需要运行
SHOW PROFILE ALL FOR QUERY n
因此我们可以查看查询是否受CPU或磁盘I/O限制.
So we can see if the query is CPU or Disk I/O bounded.
我建议使用UNION ALL或使用分治策略来分隔查询
这篇关于查询统计占用了查询时间的99%的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!