MySQL structure help for joins ( large tables)

I currently have 2 tables that are used for a select query with a simple join. The first table houses around 6-9 million rows, and this gets used as the join. The primary table is anywhere from 1mil to 300mil rows. However, I notice when I join above 10mil rows on the primary table the select query goes from instant to very slow (3+ seconds and grows).

Here is my table structure and queries.

CREATE TABLE IF NOT EXISTS `links` (
  `link_id` int(10) unsigned NOT NULL,
  `domain_id` mediumint(7) unsigned NOT NULL,
  `parent_id` int(11) unsigned DEFAULT NULL,
  `hash` int(10) unsigned NOT NULL,
  `url` text NOT NULL,
  `type` enum('html','pdf') DEFAULT NULL,
  `processed` enum('N','Y') NOT NULL DEFAULT 'N',
  UNIQUE KEY `hash` (`hash`),
  KEY `idx_processed` (`processed`),
  KEY `domain_id` (`domain_id`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8 ROW_FORMAT=COMPACT;


CREATE TABLE IF NOT EXISTS `domains` (
  `domain_id` mediumint(7) unsigned NOT NULL AUTO_INCREMENT,
  `name` varchar(170) NOT NULL,
  `blocked` enum('N','Y') NOT NULL DEFAULT 'N',
  `count` mediumint(6) NOT NULL DEFAULT '0',
  `mcount` mediumint(3) NOT NULL,
开发者_StackOverflow中文版  PRIMARY KEY (`domain_id`),
  KEY `name` (`name`),
  KEY `blocked` (`blocked`),
  KEY `mcount` (`mcount`),
  KEY `count` (`count`)
) ENGINE=MyISAM  DEFAULT CHARSET=utf8 AUTO_INCREMENT=10834389 ;

Query:

(SELECT link_id, url, hash FROM links, domains WHERE links.domain_id = domains.domain_id and mcount > 1 and processed='N' limit 200) 
UNION
(SELECT link_id, url, hash FROM links where processed='N' and type='html' limit 200)

Explain select:

| id | select_type  | table      | type  | possible_keys           | key      |     key_len | ref                       | rows    | Extra       |
+----+--------------+------------+-------+-------------------------+---------------        +---------+---------------------------+---------+-------------+
|  1 | PRIMARY      | domains    | range | PRIMARY,mcount          | mcount        | 3       | NULL                      |  257673 | Using where | 
|  1 | PRIMARY      | links      | ref   | idx_processed,domain_id | domain_id     | 3       | crawler.domains.domain_id |       1 | Using where | 
|  2 | UNION        | links      | ref   | idx_processed           | idx_processed | 1       | const                     | 7090017 | Using where | 
| NULL | UNION RESULT | <union1,2> | ALL   | NULL                    | NULL          | NULL    | NULL                      |    NULL |             | 
+----+--------------+------------+-------+-------------------------+---------------+---------+---------------------------+---------+-------------+

Right now, I'm trying a partition with 20 partitions on links using domain_id as the key.

Any other options would be greatly appreciated.

---

A single SELECT statement would replace your entire UNION statement:

SELECT link_id, url, hash
FROM links, domains 
WHERE links.domain_id = domains.domain_id 
      AND mcount > 1
      AND processed='N' 
      AND type='html'

This may not be THE answer you are looking for, but it should help you simplify your question.

---

When things suddenly slow down you might want to check the size of your indexes (used in the query execution) vs size of various mysql buffers.