帮助加快 MySql 查询

Question

如何加快这个 mysql 查询的速度？

   SET @rank = 0;  

  UPDATE dbTable 
     SET rank_fd = @rank := @rank + 1 
ORDER BY fd2 DESC, fd3 DESC;

此查询更新约 300,000 个表行。它只是花费了太长时间......简单的控制面板（涡轮面板？）说mysqld只使用8-9％的cpu。我知道这总是正确的，但是这个查询在我的 Core i7 机器上花费的时间几乎没有。我认为在我的 P4 专用服务器上大约需要 5-10 分钟......这不会是一个问题，除非我必须为不同的值多次执行此操作。它还使其他 mysql 操作极其缓慢（几乎挂起），从而严重影响网站性能。

软件：

• CentOS 5.4 版（最终版），
• 简单控制面板版本 1.4.8，
• PHP 5.2.16
• MySQL 客户端版本：5.1.52

硬件：

• 英特尔奔腾 4 CPU 3.0GHz
• 2GB 内存

更多信息（添加于 2011 年 3 月 6 日 @ 10:06PM CST）：

//Core i7 920 @ 2.6GHz, 6GB Ram
UPDATE dbTable SET rank_fd =999999999;
#275037 row(s) affected. ( Query took 7.0708 sec )

SET @rank =0;
#Your SQL query has been executed successfully ( Query took 0.0003 sec )

UPDATE dbTable SET rank_fd = @rank := @rank +1 ORDER BY fd2 DESC, fd3 DESC ;
#275037 row(s) affected. ( Query took 9.9931 sec )



//P4 3.0GHz, 2GB Ram
UPDATE dbTable SET rank_fd =999999999;
#Affected rows: 291468 (Query took 8.2165 sec)

SET @rank =0;
#Your SQL query has been executed successfully (Query took 0.0002 sec)

UPDATE dbTable SET rank_fd = @rank := @rank +1 ORDER BY fd2 DESC, fd3 DESC ;
#Affected rows: 291469 (Query took 305.2104 sec)  

更多信息（添加于 2011 年 3 月 7 日 @ 下午 6:37 CST）：

我有一些新信息。如果我在 P4 上做一个 select 语句：

SET @rank =0;
SELECT @rank := @rank +1 AS rank_fd FROM dbTable ORDER BY fd2 DESC, fd3 DESC LIMIT 0, 300000;  
#Showing rows 0 - 29 (292,437 total, Query took 3.0448 sec)

计算一切只需要 3 秒。 no-calc 批量更新语句只需要 8 秒。什么是所有额外的工作导致它在我原来的陈述中超过 300 秒。有没有办法在不涉及 PHP 的 select calc 语句之后捕获更新。我之所以这么说，是因为如果我在 PHP 中循环它，它会比原来的语句花费更长的时间。

感谢到目前为止的所有帮助！！！

Answer 1

您正在修改表格中的每一行，当然会很慢。根据rank_fd 的使用方式，您可能只需在fd2 和fd3 上放置索引并在运行时计算排名即可。

Answer 2

您可以使用存储过程和临时表来加快排名过程。

以下示例使用包含 150 万行的玩家得分表，其中 500K 玩家得分超过 3 轮 (500K * 3)，并在大约 50 分钟内更新第 1 轮（500K 行）的排名。 5 秒。

希望这会有所帮助:)

示例表和存储过程

drop table if exists player_scores;
create table player_scores
(
round_id smallint unsigned not null,
player_id int unsigned not null,
score_1 int unsigned not null default 0,
score_2 int unsigned not null default 0,
rank int unsigned not null default 0,
primary key (round_id, player_id)
)
engine=myisam;

drop procedure if exists update_player_score_ranking;

delimiter #
create procedure update_player_score_ranking
(
p_round_id smallint unsigned
)
begin

  create table tmp_player_scores engine=memory select 
    round_id, player_id, score_1, score_2, @rank:= @rank + 1 as rank
  from 
    player_scores 
  inner join (select @rank:=0) r
  where
   round_id = p_round_id
  order by 
    score_1 desc, score_2 desc;

  delete from player_scores where round_id = p_round_id;
  insert into player_scores select * from tmp_player_scores;

  drop table if exists tmp_player_scores;

end #

delimiter ;

测试结果：

select count(*) from player_scores;
+----------+
| count(*) |
+----------+
|  1500000 |
+----------+
1 row in set (0.00 sec)

select count(*) from player_scores where round_id = 1;
+----------+
| count(*) |
+----------+
|   500000 |
+----------+
1 row in set (0.07 sec)

select * from player_scores where round_id = 1 order by score_1 desc, score_2 desc limit 5;
+----------+-----------+---------+---------+------+
| round_id | player_id | score_1 | score_2 | rank |
+----------+-----------+---------+---------+------+
|        1 |    456937 |   65534 |   49579 |    0 |
|        1 |     72439 |   65534 |   44537 |    0 |
|        1 |     16427 |   65534 |   43045 |    0 |
|        1 |    259871 |   65534 |   32095 |    0 |
|        1 |    324702 |   65534 |   15227 |    0 |
+----------+-----------+---------+---------+------+
5 rows in set (0.71 sec)


call update_player_score_ranking(1);
Query OK, 0 rows affected (5.57 sec)

select * from player_scores where round_id = 1 order by rank limit 5;
+----------+-----------+---------+---------+------+
| round_id | player_id | score_1 | score_2 | rank |
+----------+-----------+---------+---------+------+
|        1 |    456937 |   65534 |   49579 |    1 |
|        1 |     72439 |   65534 |   44537 |    2 |
|        1 |     16427 |   65534 |   43045 |    3 |
|        1 |    259871 |   65534 |   32095 |    4 |
|        1 |    324702 |   65534 |   15227 |    5 |
+----------+-----------+---------+---------+------+
5 rows in set (1.29 sec)