JOIN查询非常慢

Question

I've got a query:

EXPLAIN 
SELECT i.ipStart, i.ipEnd, i.descr 
FROM orgs i 
JOIN visit_count v 
on inet_aton(v.ipaddress) BETWEEN i.ipStart and i.ipEnd
WHERE v.last_visit BETWEEN '2014-10-10' AND '2014-10-22' 
AND v.user_id = 1

The above is extremely slow as I increase the time period. The above takes about 20 seconds.

My two table schemas:

`visit_count` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `visitCount` int(11) NOT NULL,
  `user_id` int(11) NOT NULL,
  `cookieId` varchar(50) NOT NULL,
  `ipaddress` varchar(50) NOT NULL,
  `last_visit` datetime NOT NULL,
  PRIMARY KEY (`id`),
  KEY `cookieId` (`cookieId`),
  KEY `user_id` (`user_id`),
  KEY `last_visit` (`last_visit`),
  KEY `user_id_2` (`user_id`,`last_visit`)
) ENGINE=MyISAM DEFAULT CHARSET=latin1;


`orgs` (
  `id` int(11) unsigned NOT NULL AUTO_INCREMENT,
  `ipStart` bigint(100) DEFAULT NULL,
  `ipEnd` bigint(100) DEFAULT NULL,
  `land` varchar(250) DEFAULT NULL,
  `descr` varchar(250) DEFAULT NULL,
  `ipStartRead` varchar(250) DEFAULT NULL,
  `ipEndRead` varchar(250) DEFAULT NULL,
  PRIMARY KEY (`id`),
  KEY `ipStart` (`ipStart`,`ipEnd`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;

visit_count has about 1,6 mil rows

and

orgs has about 14k rows

Any ideas how I could make this go faster?

Edit

The explain:

+----+-------------+-------+-------+------------------------------+-----------+---------+------+-------+---------------------------------------------------+
| id | select_type | table | type  | possible_keys                | key       | key_len | ref  | rows  | Extra                                             |
+----+-------------+-------+-------+------------------------------+-----------+---------+------+-------+---------------------------------------------------+
   1   SIMPLE        v       range   user_id,last_visit,user_id_2   user_id_2         9   NULL     357   Using index condition
   1   SIMPLE        i       ALL     NULL                           NULL        NULL      NULL   22068   Using where; Using join buffer (Block Nested Loop)

Answer 1

Doing address lookups like this is very innefficient. Firstly, you're using a function as a join expression which means you can't use an index, and range queries (particularly on IP addresses) are very innefficient.

Using the same data type in both tables would help. To solve the range lookup problem there are 2 solutions:

1) instead of having a reference set of arbitrary network ranges, used fixed ranges (eg 255.255.255.0 netmasked) This could be in the form of an:m decomposition between the current tables which would also allow you to keep the differing data types but use indexes.

This a rather kludgy solution.

2) Keep the reference set of IP addresses in a geospatial table. Note that although the address range is a 1-dimensional space, the MySQL Geospatial database only really understands 2-d spaces hence you need to fudge the space a little - this is nicely described here

Note that the time range lookup is less affected as the upper and lower bounds of the predicates are bother defined by the same column in the table.