mysql查询优化

Posted 2023-04-14

【中文标题】mysql查询优化

【英文标题】：mysql query optimization

【发布时间】：2012-10-26 04:32:36

【问题描述】：

我的任务是帮助加快此查询，我认为这些表上的某些索引设置不正确。我也相信它们不会全部用于 b/c 一个函数应用于我有索引的 col。谁能看到我如何优化这些表或查询？ requests 表将是 3 个表中最大的表，将有超过 20 万条记录。 devices 目前有大约 500 条记录，clients 也会更小。

查询：

explain extended SELECT MAX(Request.datetime) AS datetime, Device.id,
       Device.client_id, Device.mac_address, Device.type, Device.manufacturer,
       Device.model_number, Client.id, Client.email_address,
       Request.device_id, Request.datetime, Request.ip_address
  FROM livefi.devices AS Device
  LEFT JOIN livefi.clients AS Client
    ON (Client.id         = Device.client_id)
 INNER JOIN livefi.requests AS Request
    ON (Request.device_id = Device.id)
 GROUP BY Request.device_id, Request.client_id

+----+-------------+---------+--------+---------------------------------------------------------+---------------+---------+-------------------------+------+----------+---------------------------------+
| id | select_type | table   | type   | possible_keys                                           | key           | key_len | ref                     | rows | filtered | Extra                           |
+----+-------------+---------+--------+---------------------------------------------------------+---------------+---------+-------------------------+------+----------+---------------------------------+
|  1 | SIMPLE      | Device  | ALL    | PRIMARY                                                 | NULL          | NULL    | NULL                    |  617 |   100.00 | Using temporary; Using filesort |
|  1 | SIMPLE      | Client  | eq_ref | PRIMARY                                                 | PRIMARY       | 4       | livefi.Device.client_id |    1 |   100.00 |                                 |
|  1 | SIMPLE      | Request | ref    | idx_device_id,inx_requests_deviceId_datetime_ip_address | idx_device_id | 5       | livefi.Device.id        |  144 |   100.00 | Using where                     |
+----+-------------+---------+--------+---------------------------------------------------------+---------------+---------+-------------------------+------+----------+---------------------------------+
3 rows in set, 1 warning (0.04 sec)

表格：

CREATE TABLE `clients` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `email_address` varchar(100) DEFAULT NULL,
  `mac_address` varchar(17) DEFAULT NULL,
  PRIMARY KEY (`id`),
  UNIQUE KEY `email_address` (`email_address`),
  KEY `idx_mac_address` (`mac_address`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8


CREATE TABLE `devices` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `client_id` int(10) unsigned DEFAULT NULL,
  `mac_address` varchar(17) DEFAULT NULL,
  `type` varchar(25) DEFAULT NULL,
  `manufacturer` varchar(100) DEFAULT NULL,
  `model_number` varchar(50) DEFAULT NULL,
  PRIMARY KEY (`id`),
  UNIQUE KEY `mac_address` (`mac_address`),
  KEY `idx_mac_address` (`mac_address`),
  KEY `fk_devices_clients1` (`client_id`),
  CONSTRAINT `fk_devices_clients1` FOREIGN KEY (`client_id`) REFERENCES `clients` (`id`) ON DELETE NO ACTION ON UPDATE NO ACTION
) ENGINE=InnoDB AUTO_INCREMENT=492 DEFAULT CHARSET=utf8


CREATE TABLE `requests` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `client_id` int(10) unsigned DEFAULT NULL,
  `device_id` int(10) unsigned DEFAULT NULL,
  `domain_id` int(10) unsigned DEFAULT NULL,
  `ip_address` varchar(15) DEFAULT NULL,
  `datetime` datetime DEFAULT NULL,
  `gmt_offset` time DEFAULT NULL,
  `request_method` varchar(15) DEFAULT NULL,
  `url` text,
  `http_protocol` varchar(20) DEFAULT NULL,
  `http_status_code` varchar(20) DEFAULT NULL,
  `request_size` int(10) unsigned DEFAULT '0',
  `referer` text,
  `user_agent` text,
  `squid_cache_response` varchar(255) DEFAULT NULL,
  PRIMARY KEY (`id`),
  KEY `idx_client_id` (`client_id`),
  KEY `idx_datetime` (`datetime`),
  KEY `idx_device_id` (`device_id`),
  KEY `idx_domain_id` (`domain_id`),
  KEY `idx_id` (`id`),
  KEY `idx_request_size` (`request_size`),
  KEY `inx_requests_deviceId_datetime_ip_address` (`device_id`,`datetime`,`ip_address`),
  CONSTRAINT `fk_requests_clients` FOREIGN KEY (`client_id`) REFERENCES `clients` (`id`) ON DELETE NO ACTION ON UPDATE NO ACTION,
  CONSTRAINT `fk_requests_devices1` FOREIGN KEY (`device_id`) REFERENCES `devices` (`id`) ON DELETE NO ACTION ON UPDATE NO ACTION,
  CONSTRAINT `fk_requests_domains1` FOREIGN KEY (`domain_id`) REFERENCES `domains` (`id`) ON DELETE NO ACTION ON UPDATE NO ACTION
) ENGINE=InnoDB AUTO_INCREMENT=200523 DEFAULT CHARSET=utf8

【参考方案1】：

我建议：

ALTER TABLE requests ADD INDEX (device_id, client_id, datetime);

另外，请注意，您不应在 SELECT 子句中包含 Request.datetime 或 Request.ip_address，因为它们不是 GROUP BY 子句的一部分。这是因为requests 中的几行可能具有相同的device_id 和client_id，而为datetime 或ip_address 选择哪个值有些随机。

【参考方案2】：

我会在 ( Device_ID, ID ) 上的设备上有一个索引，我过去曾提供过这个索引，并且显然对其他人来说效果很好。日期可能是查询中的一个瓶颈，但整数很容易。

如果您仅对 Device AND MAX(ID) 进行预查询，这实质上将返回最新的日期/时间...由于 ID 是自动递增的，因此其日期/时间值将直接关联为最近的.. 所以它们是一回事。

一个例外情况是，如果您处于多台服务器的某种集群环境中，可能会添加自己的 auto-inc ID 列，这可能会导致错误的最大条目。如果是这种情况，则将索引更改为 (Device_ID, DateTime)。

也就是说，我会执行以下操作

SELECT STRAIGHT_JOIN
      PerQuery.Device_ID,
      PreQuery.LastRequestID as Request_ID,
      RDtl.datetime, 
      d.client_id, 
      d.mac_address, 
      d.type, 
      d.manufacturer,
      d.model_number, 
      c.id, 
      c.email_address,
      r.ip_address
   FROM 
      ( select rq.Device_ID,
               MAX( rq.ID ) as LastRequestID
            from
               livefi.Requests rq
            group by
               rq.Device_ID ) PreQuery
         JOIN livefi.Request RDtl
            ON PreQuery.LastRequestID = RDtl.ID
            JOIN livefi.Clients c
               ON RDtl.ClientID = c.ID
         JOIN livefi.Devices d
               ON PreQuery.Device_ID = d.ID

预查询是唯一具有 group by 的查询，并且会直接优化到索引...由于每个设备只会返回一条记录及其对应的“请求”ID，因此只会加入原始请求表“其他”详细信息、客户端和设备详细信息。