Question

无论如何都可以从中获得更好的表现。

SELECT fdyofmth, SUM(total_restored) AS total_restored
FROM sum_restored
WHERE fdyofmth BETWEEN '2010-11-01' AND '2011-11-01'
GROUP BY fdyofmth
ORDER BY fdyofmth DESC

结果需要3秒。

+------------+----------------+
| fdyofmth   | total_restored |
+------------+----------------+
| 2011-11-01 |          39793 |
| 2011-10-01 |          52367 |
| 2011-09-01 |          52141 |
| 2011-08-01 |          56515 |
| 2011-07-01 |          54668 |
| 2011-06-01 |          54874 |
| 2011-05-01 |          60025 |
| 2011-04-01 |          60005 |
| 2011-03-01 |          63856 |
| 2011-02-01 |          64472 |
| 2011-01-01 |          58943 |
| 2010-12-01 |          57005 |
| 2010-11-01 |          55731 |
+------------+----------------+

EXPLAIN结果 -

+----+-------------+--------------+------+---------------+------+---------+------+--------+----------------------------------------------+
| id | select_type | table        | type | possible_keys | key  | key_len | ref  | rows   | Extra                                        |
+----+-------------+--------------+------+---------------+------+---------+------+--------+----------------------------------------------+
|  1 | SIMPLE      | sum_restored | ALL  | NULL          | NULL | NULL    | NULL | 752660 | Using where; Using temporary; Using filesort |
+----+-------------+--------------+------+---------------+------+---------+------+--------+----------------------------------------------+
1 row in set

我试过把一个fdyofmth的钥匙实际上放慢了两倍。

非常感谢任何帮助。

谢谢。

Answer 1

我认为它很慢，因为你有这么多的数据，你的sql必须遍历所有数据（不从整个数据集中选择一个），所以index在这里不能帮助。

我建议的选项是使用并行性：

按照@newtover
的建议添加索引
使用12个线程来检索每个月的结果。

甚至更简单：

只是将结果预先计算到另一个表中，当查询时，从该表中查询。

Answer 2

试试这个：如果日期不一样，那么

SELECT fdyofmth, SUM(total_restored) AS total_restored
FROM sum_restored
WHERE fdyofmth >= '2010-11-01' AND fdyofmth <= '2011-12-01'
GROUP BY fdyofmth
ORDER BY fdyofmth DESC

希望有所帮助

Answer 3

根据您的EXPLAIN结果，查询不使用任何索引。您需要在fdyofmth上设置索引：

ALTER TABLE sum_restored ADD INDEX fdyofmth_index (fdyofmth);

或另一个建议=）：

SELECT
    SUM (IF(fdyofmth = '2010-11-01', total_restored, NULL)) as `2010-11-01`,
    SUM (IF(fdyofmth = '2010-12-01', total_restored, NULL)) as `2010-12-01`,
    SUM (IF(fdyofmth = '2011-01-01', total_restored, NULL)) as `2011-01-01`,
    SUM (IF(fdyofmth = '2011-02-01', total_restored, NULL)) as `2011-02-01`,
    SUM (IF(fdyofmth = '2011-03-01', total_restored, NULL)) as `2011-03-01`,
    SUM (IF(fdyofmth = '2011-04-01', total_restored, NULL)) as `2011-04-01`,
    SUM (IF(fdyofmth = '2011-05-01', total_restored, NULL)) as `2011-05-01`,
    SUM (IF(fdyofmth = '2011-06-01', total_restored, NULL)) as `2011-06-01`,
    SUM (IF(fdyofmth = '2011-07-01', total_restored, NULL)) as `2011-07-01`,
    SUM (IF(fdyofmth = '2011-08-01', total_restored, NULL)) as `2011-08-01`,
    SUM (IF(fdyofmth = '2011-09-01', total_restored, NULL)) as `2011-09-01`,
    SUM (IF(fdyofmth = '2011-10-01', total_restored, NULL)) as `2011-10-01`,
    SUM (IF(fdyofmth = '2011-11-01', total_restored, NULL)) as `2011-11-01`
FROM sum_restored;

Answer 4

您应该解决此外部查询更改。

创建另一个表格以保存每个过去一个月的现成计算摘要
减少了很多桌子的大小。每年或每月或每行的另一个表格更改。

简单查询慢

4 个答案: