更好的性能，JOIN 还是多个查询？

Question

Whats the best way to achieve best performance for getting data from multiple table?

I have these following tables

Applicants ( 50,101 rows )
-id
-first_name
-email

Phones ( 50,151 rows )
-id
-number
-model_id
-model_type

Address (100,263 rows)
-id
-state
-model_id
-model_type

Business (26 rows)
-id
-company
-model_id
-model_type

My desired result

 id | first_name | email | number | company | state
----+------------+-------+--------+---------+------
  1 |    test    |   -   |   -    |    -    |   -

Im using SQLyog to perform this query below and its very slow, I have thousands of data on these tables

SELECT `app`.`id`,`app`.`first_name`, `app`.`email`, `p`.`number`, `b`.`company`, `add`.`state`
FROM `applicants` AS `app`
LEFT JOIN  phones AS `p` ON `app`.`id` = `p`.`model_id` 
    AND `p`.`model_type` = 'App\\Models\\Applicant'
LEFT JOIN `businesses` AS `b` ON `app`.`id` = `b`.`model_id` 
    AND `b`.`model_type` = 'App\\Models\\Applicant'
LEFT JOIN `addresses` AS `add` ON `app`.`id` = `add`.`model_id` 
    AND `b`.`model_type` = 'App\\Models\\Applicant'
LIMIT 10

summary, takes 25.794 to finish

Execution Time : 25.792 sec
Transfer Time  : 0.001 sec
Total Time     : 25.794 sec

What would be the best way to achieve my goal? like should a perform a separate multiple query for each phone, business and address? though Im not sure how to achieve my desired result with multiple query

Answer 1

It really depends on the specific situation what will be faster. Also you can probably optimize your situation by creating the proper indexes. For example if you query a lot by model_id and model_type , you could create an index on either, or both of the fields.

I would suggest running the query with joins with an EXPLAIN in front, ie EXPLAIN SELECT {your query} . That will give you some insights on how MySQL executes your query. Then you can try the same with separate queries. Then add indexes and see if they are used. Then choose the best performing solution.

About indexes:

Introduction: https://www.mysqltutorial.org/mysql-index/mysql-create-index/

More in-depth: https://use-the-index-luke.com/

Answer 2

From my understanding, there is no general answer on this question. One of the possible solutions is denormalisation. Creating extra table with all the data you need periodically. It helps a lot in some cases but, unfortunately, just not possible to do this way in other cases.

Answer 3

Here are the things to think about in this case. Arguments for a single query:

• Databases are designed to handle complicated queries, so the JOIN s are probably faster in the database.
• Each query incurs overhead for moving the query into the database and data out of the database.

In favor of a multiple queries:

• Query optimizers are not perfect, so they might come up with the wrong plan.
• Returning data as a single result set often requires a "wide" format for the data with many repeated columns. Returning more data is slower.

In general, the balance is on the single query, but that is not always true. For instance, if the database is fast but the bandwidth to the application is slow, the last bullet may consistently be the dominating factor.

I am surprised that your query takes so much time. You don't have an ORDER BY or GROUP BY so the time to the first result should be pretty fast. You might be able to have it run faster by simply doing a subselect on app :

FROM (SELECT app.* 
      FROM `applicants` `app`
      LIMIT 10
     ) app . . . 

Answer 4

Joining tables is taking all of the data from each table and combining it into one. It is expected to have a slower loading time because of this. If you are only gathering specific data from each table then running seperate queries might be the better idea. It depends on how your application is laid out and how you are using this information.