MySQL中处理慢sql时，如果看完执行计划之后还拿不定优化的方案的时候，可能需要借助profile的来看一看具体慢在哪里，整个过程如下：

mysql> set profiling=1;
Query OK, 0 rows affected (0.00 sec)

mysql> select now();
+---------------------+
| now()               |
+---------------------+
| 2016-07-11 11:10:37 |
+---------------------+
1 row in set (0.00 sec)

mysql> show profiles;
+----------+------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Query_ID | Duration   | Query                                                                                                                                                                                                                                                                                                        |
+----------+------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|        1 | 4.79560350 | SELECT rgu.user_id as recipient_id, 0 as recipient_type, srg.alarm_settings_id, srg.recipient_group_id, rg.notification_type, u.agreement_id, u.name, u.email, u.mobile_no, u.wechat_no, u.notification_hour_from, u.notification_hour_to
FROM NL_U_ALARM_SETTINGS_RECIPIENT_GROUP srg 
INNER JOIN NL_U_ALAR |
+----------+------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.02 sec)

mysql> set @query_id=1;
Query OK, 0 rows affected (0.00 sec)

mysql> select state,sum(duration) as Total_R, round(100*sum(duration)/(select sum(duration) from information_schema.profiling where query_id=@query_id),2) as Pct_R, count(*) as Calls, sum(duration)/count(*) as "R/Call" from information_schema.profiling where query_id=@query_id group by state order by Total_R desc;
+----------------------+----------+-------+-------+--------------+
| state                | Total_R  | Pct_R | Calls | R/Call       |
+----------------------+----------+-------+-------+--------------+
| Sending data         | 4.753893 | 99.13 |     4 | 1.1884732500 |
| removing tmp table   | 0.038097 |  0.79 |     1 | 0.0380970000 |
| System lock          | 0.000906 |  0.02 |     1 | 0.0009060000 |
| Opening tables       | 0.000517 |  0.01 |     1 | 0.0005170000 |
| checking permissions | 0.000354 |  0.01 |     8 | 0.0000442500 |
| preparing            | 0.000332 |  0.01 |     3 | 0.0001106667 |
| optimizing           | 0.000276 |  0.01 |     3 | 0.0000920000 |
| starting             | 0.000262 |  0.01 |     1 | 0.0002620000 |
| executing            | 0.000220 |  0.00 |     3 | 0.0000733333 |
| statistics           | 0.000212 |  0.00 |     3 | 0.0000706667 |
| closing tables       | 0.000185 |  0.00 |     1 | 0.0001850000 |
| freeing items        | 0.000164 |  0.00 |     1 | 0.0001640000 |
| cleaning up          | 0.000128 |  0.00 |     1 | 0.0001280000 |
| logging slow query   | 0.000045 |  0.00 |     1 | 0.0000450000 |
| query end            | 0.000013 |  0.00 |     1 | 0.0000130000 |
+----------------------+----------+-------+-------+--------------+
15 rows in set (0.01 sec)

mysql> set profiling=0;
Query OK, 0 rows affected (0.00 sec)