提高MySQL查询的性能

Question

I have a MySQL query that is currently running at 2.5 seconds and I want to trim it down to 2 seconds max.

Here is the query:

SELECT SQL_CALC_FOUND_ROWS Distinct(c.id) 
FROM   `content` c 
       INNER JOIN `actions` AS action 
               ON c.parent_id = action.primary_id 
WHERE  1 = 1 
       AND c.site_id IN ( 1, 2 ) 
       AND c.type IN ( 'topic' ) 
       AND c.status = 'visible' 
       AND ( c.lock = 0 
              OR c.site_id = 1 ) 
       AND ( c.level IN ( 0 ) 
              OR ( c.level IN ( 10 ) 
                   AND action.user_id = 123 
                   AND action.type IN ( 'group_follow' ) ) ) 
ORDER  BY c.date_updated DESC 
LIMIT  20 

Here are the current stats:

• table content has 55k rows
• table actions has 87k rows

For content, I have indexes on:

• parent_id (parent_id)
• user_id (user_id)
• type_status_date (type,status, date)
• type_status_updateddate (type, status, date_updated)
• site_id (site_id)
• type_status_level (type, status, level)
• type_parentid_level (type, parent_id, level)

For actions, I have indexes on:

• primary_id (primary_id)
• site_id (site_id)
• type (type)

Any help and suggestions would be much appreciated.

Thanks all!

Answer 1

Since you're doing a "ORDER BY c.date_updated DESC LIMIT 20", you probably want an index on c.date_updated. That way, mysql can scan that table first in reverse date_updated order and stop as soon as it gets the 20 rows.

You should definitely use EXPLAIN to check your query before and after the change, to see if the optimizer selects that order. There are ways to force the use of the date_updated index if the optimizer doesn't select it naturally.

Answer 2

in where clause you must have order. firsts must be a column that have a index.

for example for index type_status_updateddate (type, status, date_updated)

if we write

where status = 'blabla' and type = '1'

index not using nut

where type='1' and status='blala'

use index.

mysql use index if column is left in index(in multi column index). for example for index type_status_updateddate (type, status, date_updated)

if we write where status = 'blablabla' index not using

this fact do using multicolumn indexes not cool

and

c.type IN ( 'topic' )

may be replace on

c.type = 'topic' 

sorry for my english.