optimize mysql query across multiple tables - mysql

Question

I am having some issues optimizing the following query that joins multiple tables.
SELECT count(*) as count, `snapshot_id`, `snapshot_guid`, `image`, `subject`, `name`, `brands`.`facebook`, `brands`.`brand_id`, `brand_guid`, `date_sent`
FROM (`snapshots`)
INNER JOIN `brands` ON `snapshots`.`brand_id` = `brands`.`brand_id`
WHERE `snapshots`.`status` = 1
AND `brands`.`status` = 1
AND `brands`.`archive` = 0
GROUP BY `snapshots`.`brand_id`, `snapshots`.`subject`
ORDER BY `date_sent` desc
LIMIT 20
Execution Time: 4.9 seconds
Using EXPLAIN:
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE brands ALL PRIMARY,brand_id,status,brand_status NULL NULL NULL 338 Using where; Using temporary; Using filesort
1 SIMPLE snapshots ref brand_id,status,snapshot_brand_status snapshot_brand_status 5 mockd_catalog.brands.brand_id,const 166
Describe Tables for Brands and Snapshots:
Describe brands
Field Type Null Key Default Extra
brand_id int(11) NO PRI NULL auto_increment
brand_guid char(12) NO MUL NULL
friendly varchar(128) YES NULL
name varchar(128) NO NULL
url varchar(2048) YES NULL
logo text YES NULL
cover text YES NULL
facebook varchar(2048) YES NULL
address_1 varchar(128) YES NULL
address_2 varchar(128) YES NULL
city varchar(50) YES NULL
state varchar(50) YES NULL
postal varchar(20) YES NULL
country_code varchar(128) YES NULL
date_created datetime YES NULL
date_modified datetime YES NULL
hp_snapshot tinyint(1) NO 0
status tinyint(1) YES MUL 0
archive tinyint(1) NO 0
Describe snapshots
Field Type Null Key Default Extra
snapshot_id int(11) NO PRI NULL auto_increment
snapshot_guid char(36) YES MUL NULL
brand_id int(11) NO MUL 0
email varchar(256) NO MUL NULL
seed_email varchar(256) NO NULL
date_sent datetime NO MUL NULL
date_created datetime NO NULL
date_modified datetime YES NULL
content_type varchar(10) YES NULL
subject varchar(256) NO NULL
source longtext YES NULL
html longtext NO NULL
html_error text YES NULL
thumbnail text YES NULL
image text YES NULL
status tinyint(1) NO MUL 0
archive tinyint(1) NO 0
tags text YES NULL
I've tried everything that I can think of, and can't seem to further optimize this query. Any help is appreciated.
Rick
edit 3/22/2015 # 10:27am ET:
CREATE TABLE `snapshots` (
`snapshot_id` int(11) NOT NULL AUTO_INCREMENT,
`snapshot_guid` char(36) DEFAULT NULL,
`brand_id` int(11) NOT NULL DEFAULT '0',
`email` varchar(256) NOT NULL,
`seed_email` varchar(256) NOT NULL,
`date_sent` datetime NOT NULL,
`date_created` datetime NOT NULL,
`date_modified` datetime DEFAULT NULL,
`content_type` varchar(10) DEFAULT NULL,
`subject` varchar(256) NOT NULL,
`source` longtext,
`html` longtext NOT NULL,
`html_error` text,
`thumbnail` text,
`image` text,
`status` tinyint(1) NOT NULL DEFAULT '0' COMMENT '-1= error, 0 = new, 1 = approved, 2 = review ',
`archive` tinyint(1) NOT NULL DEFAULT '0',
`tags` text,
PRIMARY KEY (`snapshot_id`),
KEY `snapshot_id` (`snapshot_id`) USING BTREE,
KEY `brand_id` (`brand_id`),
KEY `email` (`email`(255)),
KEY `status` (`status`),
KEY `snapshot_guid` (`snapshot_guid`) USING BTREE,
KEY `subject` (`subject`(255)),
KEY `archive` (`archive`),
KEY `archive_status` (`archive`,`status`),
KEY `date_sent` (`date_sent`) USING BTREE,
KEY `recent_snapshots` (`snapshot_id`,`snapshot_guid`,`archive`,`status`,`brand_id`,`date_sent`,`subject`(255)) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=95002 DEFAULT CHARSET=utf8;
Brands Table:
CREATE TABLE `brands` (
`brand_id` int(11) NOT NULL AUTO_INCREMENT,
`brand_guid` char(12) NOT NULL,
`friendly` varchar(128) DEFAULT NULL,
`name` varchar(128) NOT NULL,
`url` varchar(2048) DEFAULT NULL,
`logo` text,
`cover` text,
`facebook` varchar(2048) DEFAULT NULL,
`address_1` varchar(128) DEFAULT NULL,
`address_2` varchar(128) DEFAULT NULL,
`city` varchar(50) DEFAULT NULL,
`state` varchar(50) DEFAULT NULL,
`postal` varchar(20) DEFAULT NULL,
`country_code` varchar(128) DEFAULT NULL,
`date_created` datetime DEFAULT NULL,
`date_modified` datetime DEFAULT NULL,
`hp_snapshot` tinyint(1) NOT NULL DEFAULT '0',
`status` tinyint(1) DEFAULT '0',
`archive` tinyint(1) NOT NULL DEFAULT '0',
PRIMARY KEY (`brand_id`),
KEY `brand_guid` (`brand_guid`) USING BTREE,
KEY `brand_id` (`brand_id`),
KEY `status` (`status`) USING BTREE,
KEY `archive_status` (`archive`,`status`),
KEY `archive` (`archive`)
) ENGINE=InnoDB AUTO_INCREMENT=423 DEFAULT CHARSET=utf8;
The select statement:
SELECT
COUNT(*) AS count,
`snapshot_id`,
`snapshot_guid`,
`image`,
`subject`,
`name`,
`brands`.`facebook`,
`brands`.`brand_id`,
`brand_guid`,
`date_sent`
FROM
(`snapshots`)
INNER JOIN
`brands` ON `snapshots`.`brand_id` = `brands`.`brand_id`
WHERE
`snapshots`.`archive` = 0
AND `snapshots`.`status` = 1
AND `brands`.`archive` = 0
AND `brands`.`status` = 1
GROUP BY `snapshots`.`brand_id` , `snapshots`.`subject`
ORDER BY `date_sent` DESC
LIMIT 20;
Explain:
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE snapshots ref brand_id,status,archive,archive_status status 1 const 48304 Using where; Using temporary; Using filesort
1 SIMPLE brands eq_ref PRIMARY,brand_id,status,archive_status,archive PRIMARY 4 mockd_catalog.snapshots.brand_id 1 Using where

Answer 1

Ther are several problems :
The key on date_send is defined using hash while you are using it in order by date_sent.
You have a where clause with brand.status and brand.archive. archive is only in a combined key brand_id, status, archive, but it cannot be used because the first column (brand_id) is not used in the where clause. Create an index for archive only, or a composite (status, archive).
subject needs an index on its own. Currently its index is buried deep down in a composite index.
As a general rule the order of the columns in a composite indexes important. The index is only useful if the used columns are the first ones. Also they are less effective the wider the first column is. This means that
KEY `snap_indx` (`snapshot_id`,`snapshot_guid`,`email`(255),`date_sent`,`subject`(255),`status`,`archive`)
is inefficient because subject with its 255 bytes comes before status and archive which have only 4 bytes each.