Mysql 计算两个日期条件之间非常大的行的性能

Question

我有一个 table 在 Innodb 中有超过 2000 万行。

列是

id, viewable_id, visitor, viewed_at

其中 viewable_id 和 viewed_at 是索引。当我执行以下查询时

SELECT COUNT(*) 
FROM views_users 
WHERE (viewable_id = 2) 
  and (viewed_at between '2021-04-19 01:38:37' 
  and '2021-06-30 01:38:37');

=> take (3 min 6.72 sec)

解释是

| id | select_type | table       | partitions | type | possible_keys                                             | key                           | key_len | ref   | rows    | filtered | Extra       |
+----+-------------+-------------+------------+------+-----------------------------------------------------------+-------------------------------+---------+-------+---------+----------+-------------+
|  1 | SIMPLE      | views_users | NULL       | ref  | views_users_viewable_id_index,views_users_viewed_at_index | views_users_viewable_id_index | 8       | const | 9554594 |    50.00 | Using where

如何才能将性能提高到 4 秒以内？

CREATE TABLE views_users (
    id int unsigned NOT NULL AUTO_INCREMENT,
    viewable_type varchar(255) NOT NULL,
    viewable_id bigint unsigned NOT NULL,
    visitor text,
    collection varchar(255) DEFAULT NULL,
    viewed_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
    PRIMARY KEY (id),
    KEY user_id (viewable_id)
) ENGINE=InnoDB AUTO_INCREMENT=20995848
    DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci

Answer 1

对于此查询：

SELECT COUNT(*)
FROM views_users
WHERE viewable_id = 2 and
      viewed_at between '2021-04-19 01:38:37' and '2021-06-30 01:38:37';

您可以创建一个索引：

CREATE INDEX idx_views_users_viewable_id_viewed_at ON views_users(viewable_id, viewed_at);

Answer 2

我通过应用 MySQL partiotons 将性能提高到不到 2 秒。

我使用 viewed_at 列按范围分区。将 viewed_at 类型从 timestamp 更改为 datatime 并将其作为 id 的主键。使 cronjob 在每个月的第一天运行，将最后一个分区重组为另一个分区，依此类推。

Mysql 计算两个日期条件之间非常大的行的性能

Mysql count performance on very big rows between two dates conditions

mysql

sql

count

query-optimization

database-performance