MySQL的InnoDB的不是MyISAM數據

Question

我有兩個錶慢得多（降低到重要字段）：

syskeywordobjects: 
pksyskeywordobjects BIGINT 
fksyskeywords BIGINT 
fkcontents BIGINT 

fkcontents INDEX (fkcontents) 
fksyskeywords INDEX (fksyskeywords) 
fkcontents_fksyskeywords INDEX (fkcontents, fksyskeywords) 


syskeywords: 
pksyskeywords BIGINT 
keyword VARCHAR 

keyword INDEX (keyword) 

在這兩個表我運行下面的查詢：

SELECT k.pksyskeywords, k.keyword, COUNT(k.pksyskeywords) AS counter 
FROM syskeywordobjects ko INNER JOIN syskeywords k ON ko.fksyskeywords = k.pksyskeywords 
WHERE (
    k.pksyskeywords <> 1218713201167374664 
    AND EXISTS (
     SELECT innerko.pksyskeywordobjects 
     FROM syskeywordobjects innerko 
     WHERE ko.fkcontents = innerko.fkcontents 
     AND innerko.fksyskeywords = 1218713201167374664 
    ) 
) 
GROUP BY k.pksyskeywords, k.keyword 
ORDER BY counter DESC 
LIMIT 20 

如果表使用MyISAM的查詢需要大約1-2s，但如果我使用InnoDB（我必須）它需要25-30s。爲什麼InnoDB慢20倍？

如需進一步的幫助，請查看解釋結果。

InnoDB: 
id select_type   table type possible_keys          key      key_len ref     rows Extra 
1  PRIMARY    k  range PRIMARY            PRIMARY     8   NULL     52051 Using where; Using temporary; Using filesort 
1  PRIMARY    ko  ref  fksyskeywords          fksyskeywords    8   k.pksyskeywords  1  Using where 
2  DEPENDENT SUBQUERY innerko ref  fkcontents,fksyskeywords,fkcontents_fksyskeywords fkcontents_fksyskeywords 16  ko.fkcontents,const 1  Using index 

MyISAM: 
id select_type   table type possible_keys          key       key_len ref     rows  Extra 
1  PRIMARY    ko  index fksyskeywords          fkcontents_fksyskeywords 16   NULL     277823 Using where; Using index; Using temporary; Using f... 
1  PRIMARY    k  eq_ref PRIMARY            PRIMARY      8   ko.fksyskeywords  1  
2  DEPENDENT SUBQUERY innerko ref  fkcontents,fksyskeywords,fkcontents_fksyskeywords fkcontents_fksyskeywords 16   ko.fkcontents,const 1  

Answer 1

爲了讓InnoDB更快地工作，您應該始終將innodb_buffer_pool設置爲一個很大的值。它允許InnoDB將工作數據集存儲在內存中（表數據，索引數據結構），因此工作速度更快（RAM> IO的IO）。如果我沒有弄錯，默認值是8MB，我個人有80％的可用內存。