Question

我有一个SQL，执行时间超过一分钟，这对我来说太过分了，有人可以给出优化建议。

更新2 - 增加速度的一点修改的sql：

SELECT STRAIGHT_JOIN
kw.`id` ,  kw.`clientid` ,  kw.`day` ,  kw.`campaignid` ,  kw.`campaign` ,  kw.`adgroupid`,  kw.`adgroup` ,  kw.`matchtype` ,  
kw.`firstpagecpc` ,  kw.`keywordtext` ,  kw.`qualityscore` ,  kw.`maxcpc` ,  kw.`avgcpm` ,  kw.`status` ,  kw.`keywordtext` ,  
kw.`matchtype` ,

SUM( kw.`impressions` )                                                  AS impressions,
SUM(  kw.`clicks` )                                                      AS clicks,
SUM(  kw.`impressions` *  kw.`avgposition` ) / SUM(  kw.`impressions` )  AS avgposition, 
SUM(  kw.`cost` )                                                        AS cost,
(SUM(  kw.`clicks` ) / SUM(  kw.`impressions` ) *100)                    AS ctr,
(SUM(  kw.`cost` ) / SUM(  kw.`clicks` ))                                AS avgcpc,
IF(kw.maxcpc = 0,kw.maxcpc,ad.defaultmaxcpc)                             AS maxcpc

FROM  `keywordsreport`                                                   AS kw
JOIN  `adgroupreport`                                                    AS ad 
ON (kw.clientid = ad.clientid AND kw.`day` = ad.day)

WHERE  
kw.`campaignid` = '55347673'
AND kw.`clientid` = '6588826821'
AND kw.`day` BETWEEN '2011-07-01' AND '2011-07-31'
AND (kw.`adid` = '' OR kw.`adid` = '0')
GROUP BY kw.`keywordtext`, kw.`matchtype`

keywordsreport table indexes：

PRIMARY  PRIMARY 235232  id

Unique   UNIQUE  235232  adgroupid
                         day
                         keywordtext
                         matchtype
                         adid
INDEX    INDEX   16802   keywordtext
                          matchtype
                          clientid

CLIENTIDDAY  INDEX   6637 clientid
                          day

adgroupreport表索引：

PRIMARY          PRIMARY 1328238  id

CLIENTIDDAY  INDEX   6447     clientid
                                  day

更新3：使用EXPLAIN进行示例查询：

    i  select_type  table   type    possible_keys                                      key                key_len   ref     rows    Extra
    1   SIMPLE      kw      range   CLIENTID,CLIENTIDDAY,ADIDDAY,CAMPAIGNCLIENTDAY     CAMPAIGNCLIENTDAY  211   NULL    350 Using where; Using temporary; Using filesort
    1   SIMPLE      ad      range   CLIENTIDDAY                                        CLIENTIDDAY        65    NULL    42464   Using where

在每个表格中：keywordsreport和adgroupreport可能有大约1 000 000条记录。

Answer 1

看起来您需要clientid和/ campaignid上的索引。坚韧而没有更多细节。

Answer 2

由于您的where子句中没有adgroupid，因此无法使用kw表中的唯一索引。在clientid，day

创建一个索引

Answer 3

我会更进一步......客户可以拥有多个广告系列，因此您的索引应首先使用最小的条件。

我会有一个索引

( CampaignID, ClientId, Day )

然后只是为了匹配预期的优化索引，只需将WHERE子句调整为

WHERE
         kw.campaignid = '23213231232'
      AND kw.clientid = '$clientid'
      AND kw.`day` BETWEEN '2011-07-31' AND '2011-07-01'
      AND ( kw.adid = '' OR kw.adid = '0')

Answer 4

简单查询优化，

员工表它有两列，如 id Totalmarks（满分900分） 1 700 2 600 。。。。 n。

SELECT * FROM Employee WHERE Totalmarks BETWEEN 300 AND 600

执行查询需要一些时间。

但

ALTER TABLE Totalmarks ADD INDEX ( Totalmarks  );
SELECT * FROM Employee  WHERE Totalmarks BETWEEN 300 AND 600;

（我的）SQL优化问题

4 个答案: