索引mysql数据库中的所有列是否正确？

Question

My site is taking too long to load the contents and the CPU load used by mysql service reaches sometimes 200%. Checking the tables realized that the tables all columns were used as index. This is correct? It can affect performance?

My server configuration: 16GB RAM, 3.4GHz Most tables have about 25k ~ 50k lines. And 10~20 Columns.

All tables indexing are like this example. All columns as indexes

Answer 1

No, you don't index all columns. You index columns that are specifically involved in a WHERE clause, and sometimes if they're involved in an ORDER BY .

In this case you'd want an index on type :

SELECT name FROM users WHERE type='admin'

In this case you'd want an index on active,type :

SELECT name FROM users WHERE type='admin' AND active=1

In this case you might want an index on active,type,name :

SELECT name FROM users WHERE type-='admin' AND active=1 ORDER BY name LIMIT 10

The more indexes you add the slower writes will be but the faster reads will be. This is a classic trade-off. Evaluate carefully what indexes you need and apply them only if there will be a tangible benefit. Don't just slap them on because you feel like they should be there.

On super tiny tables, those with <1000 rows, indexes won't help that much because a table scan won't take that long. On anything non-trivial they're absolutely essential.

If you're having performance problems I'd suggest that your schema is the biggest obstacle, not the lack of indexes.

Answer 2

Indexing all columns is not correct. It affects performance of write operations. Each additional index requires additional time to update after write operation. Also each index costs additional space. You should index only the columns which will be used for searching.

Answer 3

No, you should not index all columns. It will be slower when you write data. It is not normal that your CPU reaches sometimes 200%. I suggest you check your SQL.

I'm sure optimizing SELECT statements is helpful.

http://dev.mysql.com/doc/refman/5.6/en/statement-optimization.html

1.Optimize your query with cache. Try to use parameter instead of functions.

(Thanks for @tadman's suggestion)

// no cache
$sql = "SELECT username FROM user WHERE signup_date >= CURDATE()";

// cache
$today = date("Y-m-d");
$sql = "SELECT username FROM user WHERE signup_date >= '$today'"); 

2.EXPLAIN you SELECT wisely. It helps you find potential performance issues.

EXPLAIN select name,phone from user where name="JakLiao";

3. Use "LIMIT N" when you only need first one row result.

SELECT * FROM user WHERE country = 'China' limit 1;

4. Use index when you need it in where clause.

//row "name" should add index.
select * from user where name LIKE 'Jak%' limit 10;

5. Use index when you join tables.

select company_name FROM users LEFT JOIN companies ON (users.state = companies.state) WHERE users.id = 123;

6. Avoid "select *"

//not suggest
SELECT * FROM user WHERE user_id = 1;
//suggest
SELECT username FROM user WHERE user_id = 1;

7. PROCEDURE ANALYSE(), MYSQL will help you analysis your database and data. 8. COUNT(1) when you want to know rows of a table.

//not suggest
SELECT count(*) FROM user;
//suggest
SELECT count(1) FROM user;