MySQL: Key/value store queries optimization - php

Question

I am seeking a help with setting up correct indexes (i've tried too many and now i am a bit lost), correct MySQL engine (MyIsam, InnoDB...) and help with my queries (JOINs, ...). Also I have headache when I am thinking that these queries should return count(*).
My times were more than 5 - 10 seconds per query without counting, but I am not sure if I can get better times for this big database.
I am trying to optimize this MySQL tables:
Items (~600k rows):
+-------+------------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------+------------------+------+-----+---------+----------------+
| id | int(11) unsigned | NO | PRI | NULL | auto_increment |
| type | varchar(255) | NO | PRI | NULL | |
+-------+------------------+------+-----+---------+----------------+
Items_Relationships (~1M rows):
+-------------+------------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+-------------+------------------+------+-----+---------+-------+
| lft_item_id | int(11) unsigned | NO | PRI | NULL | |
| rgt_item_id | int(11) unsigned | NO | PRI | NULL | |
| rel_type | varchar(255) | NO | PRI | NULL | |
+-------------+------------------+------+-----+---------+-------+
Items_Values (~4M rows):
+---------+------------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+---------+------------------+------+-----+---------+----------------+
| id | int(11) unsigned | NO | PRI | NULL | auto_increment |
| item_id | int(11) unsigned | NO | PRI | 0 | |
| name | varchar(255) | YES | MUL | NULL | |
| value | longtext | YES | | NULL | |
| lang | varchar(2) | YES | | NULL | |
+---------+------------------+------+-----+---------+----------------+
I am running basically these common queries:
1. query - All items with value "status" > 1:
SELECT `company`.`id` AS `id`,
`company`.`type` AS `type`
FROM `items` AS `company`
INNER JOIN `items_values` AS `value_name` ON (`company`.`id` = `value_name`.`item_id`)
WHERE `company`.`type` = 'company'
AND `value_name`.`name` = 'status'
AND CONVERT(`value_name`.`value`, SIGNED) > 1
GROUP BY `company`.`id`
ORDER BY `company`.`id` DESC
LIMIT 0, 30
2. query - All items with some values in relationship with other items:
SELECT `company`.`id` AS `id`,
`company`.`type` AS `type`
FROM `items` AS `company`
INNER JOIN `items_values` AS `value_status` ON (`value_status`.`item_id` = `company`.`id`)
INNER JOIN `items_relationships` AS `companies_categories` ON (`companies_categories`.`lft_item_id` = `company`.`id`)
INNER JOIN `items_values` AS `category_rgt` ON (`category_rgt`.`item_id` = `companies_categories`.`rgt_item_id`)
WHERE `company`.`type` = 'company'
AND `company`.`type` = 'company'
AND `value_status`.`name` = 'status'
AND CONVERT(`value_status`.`value`, SIGNED) >= 1
AND `category_rgt`.`name` = 'rgt'
AND (CONVERT(category_rgt.value, UNSIGNED) BETWEEN 2805 AND 4222)
AND `companies_categories`.`rel_type` = 'company_category'
GROUP BY `company`.`id`
ORDER BY `company`.`id` DESC LIMIT 10
OFFSET 0
Thx in advance!

Answer 1

If you ask about indexes, so you have almost everything you need already indexed.
I have just question about
| name | varchar(255) | YES | MUL | NULL | |
So I would prefer to set it as well as primary key.
I have only one suggestion about table structures.
If you have mixed string and numbers in your
| value | longtext | YES | | NULL | |
create another column int_value SIGNED or unsigned is even better.
And you should set that column as index as well (as soon as you need that column used as a filter and/or search criteria)
And fill that field on insert/update where it is applicable.
This modification will increase performance of query where you should not use CAST and/or CONVERT for millions of records like you do here:
AND CONVERT(`value_status`.`value`, SIGNED) >= 1
AND (CONVERT(category_rgt.value, UNSIGNED) BETWEEN 2805 AND 4222)
So I have no more comments about structure.
But I would ask you to try my query, just like an experiment if it faster than yours. Unfortunately I have no possibility to debug with any data. If you provide some sqlfiddle that will help a lot.
SELECT `company`.`id` AS `id`,
`company`.`type` AS `type`
FROM `items` AS `company`
INNER JOIN (
SELECT
item_id,
FROM items_values
WHERE name = 'status'
AND CONVERT(value, SIGNED) >= 1
) AS value_status
ON value_status.item_id = company.id
INNER JOIN
(
SELECT
lft_item_id
FROM
items_relationships
INNER JOIN (
SELECT
item_id
FROM
items_values
WHERE name = 'rgt'
AND (CONVERT(value, UNSIGNED) BETWEEN 2805 AND 4222)
) AS category_rgt
ON category_rgt.item_id = items_relationships.rgt_item_id
WHERE items_relationships.rel_type = 'company_category'
) as companies_categories
ON (`companies_categories`.`lft_item_id` = `company`.`id`)
WHERE `company`.`type` = 'company'
GROUP BY `company`.`id`
ORDER BY `company`.`id` DESC
LIMIT 10