MySQL大表性能优化

Question

我正在尝试解决此表的性能问题

+--------------+------------------+------+-----+---------+----------------+
| Field        | Type             | Null | Key | Default | Extra          |
+--------------+------------------+------+-----+---------+----------------+
| id           | int(10) unsigned | NO   | PRI | NULL    | auto_increment |
| direction_id | int(10) unsigned | NO   | MUL | NULL    |                |
| created_at   | datetime         | NO   |     | NULL    |                |
| rate         | decimal(16,6)    | NO   |     | NULL    |                |
+--------------+------------------+------+-----+---------+----------------+

其中包含大约 1 亿行

只有一个查询从该表中选择数据：

SELECT AVG(rate) AS rate, created_at 
FROM statistics 
WHERE direction_id = ? 
AND created_at BETWEEN ? AND ? 
GROUP BY created_at

direction_id 是外键，但选择性很差：

+----+-------------+------------+------------+------+---------------------------------+---------------------------------+---------+-------+-------+----------+---------------------------------------------------------------------+
| id | select_type | table      | partitions | type | possible_keys                   | key                             | key_len | ref   | rows  | filtered | Extra                                                               |
+----+-------------+------------+------------+------+---------------------------------+---------------------------------+---------+-------+-------+----------+---------------------------------------------------------------------+
|  1 | SIMPLE      | statistics | NULL       | ref  | statistics_direction_id_foreign | statistics_direction_id_foreign | 4       | const | 26254 |    11.11 | Using index condition; Using where; Using temporary; Using filesort |
+----+-------------+------------+------------+------+---------------------------------+---------------------------------+---------+-------+-------+----------+---------------------------------------------------------------------+

所以我正在寻找一种方法来解决这个问题并需要建议。 HASH(direction_id) 分区对我有帮助吗？ 如果有帮助，最好的方法是什么？

或者也许有其他方法可以解决它。

Answer 1

对于平均每日费率，您是说这个吗？

SELECT AVG(rate) AS rate, 
       DATE(created_at) 
    FROM statistics 
    WHERE direction_id = ? 
      AND created_at BETWEEN ? AND ? 
    GROUP BY DATE(created_at)

还有INDEX(direction_id, created, rate)——它既是“覆盖”又是“复合”。解释会说“使用索引”来表示“覆盖”，这表明整个查询可以只看索引的 BTree 来执行。因此，“覆盖”会带来额外的性能提升。

更改为涉及DATE(created_at) 的精美索引可能对此查询没有帮助。

PARTITIONing没有指明。

可能会显示“汇总表”。 http://mysql.rjweb.org/doc.php/summarytables

Answer 2

首先，让我们修复您的查询，使其成为有效的聚合查询。大概，你想要rate的日均值，所以：

SELECT AVG(rate) AS rate, DATE(created_at) as created_day
FROM statistics 
WHERE direction_id = ? AND created_at BETWEEN ? AND ? 
GROUP BY DATE(created_at)

然后，我建议创建以下索引：

create index idx_statistics on statistics (direction_id, created_at, rate);

在最新版本的 MySQL 中，我们还可以考虑在 date(create_at) 上使用索引。如果您可以接受以下where 子句：

WHERE direction_id = ? AND DATE(created_at) BETWEEN ? AND ? 

那么下面的索引就派上用场了：

create index idx_statistics on statistics (direction_id, (date(created_at)), rate);