CODEWITHSUNDEEP
mysqldatabaseperformanceindexing
haneulkim
haneulkim

Reputation: 4928

Performance difference in query between cmd and workbench mysql

I have two questions:

  1. As my title, which one is more efficient for running large query on large volume of data?

I've looked at MySQL doc which explains about workbench's performance in https://www.mysql.com/products/workbench/performance/ however I cannot seem to find any resource that specifically talks about efficiency difference between running query on cmd and query on workbench.

  1. How to optimize this query

    select
  r.user_id,
  k.id as kickscooter_id,
  st_astext(k.location) as location,
  k.created_at,
  k.serial_number,
  k_st.serial_number as states_serial_number,
  st_astext(k_st.gps) as gps_location,
  k_st.gps_updated_at,
  r.start_time,
  r.end_time
from kickscooters k
join rents r
  on k.id= r.kickscooter_id
join kickscooter_states_190614 k_st
  on k.serial_number = k_st.serial_number
order by r.rent_date
limit 999;

I've learned that creating an index allows mysql to sort things quickly therefore I've added index by 

ALTER TABLE `tablename` ADD INDEX `indexname` (`columnname`);

following answers from one of SO post "order by" taking too much time in mysql

As suggested in comments I've executed

analyze <my query>

since my server is MariaDB.

which gave me ERROR CODE 2013: LOST connection to server during query.

When I ran

explain <my query>

It worked and outputs:

id    select_type   table   type          possible_keys                                                  
1     SIMPLE        k_st    ALL  kickscooter_states_190614_serial_number_date_index             
1     SIMPLE        k       ref  PRIMARY,kickscooters_serial_number_unique,kickscooters_serial_number_index     
1     SIMPLE        r        ref    rents_kickscooter_id_foreign    

-table continued
/       key                       key_len      ref                        rows         extra

        null                        null       null                      192818947      Using temporary; Using filesort
   kickscooters_serial_number_unique 27     kickgoing_db.k_st.serial_number 1   
    rents_kickscooter_id_foreign     4      kickgoing_db.k.id              143

Upvotes: 2

Views: 100

Answers (2)

Rick James
Rick James

Reputation: 142443

There is no WHERE clause, but there is an ORDER BY and LIMIT involving r. So, the Optimizer would like to start with r. But... There seems to be no index starting with serial_number for k_st (or there is a type or collation conflict), so it abandoned r.

Please provide SHOW CREATE TABLE for all 3 tables.

Upvotes: 0

Madhur Bhaiya
Madhur Bhaiya

Reputation: 28854

Based on the Explain plan, optimizer is not able to use any index for ORDER BY rent. So try the following:

  1. Ensure that an index exists on the rent_date column of the rents table. This index will be used to optimize the ORDER BY clause. It can be a single column index, or a multi-column one (used in other scenarios). But, in case of multi-column one, you need to ensure that the rent column is the first column in the index order.
  2. Ensure that an index exists on the id column of the kickscooters table. Details about single-column / multi-column index remains the same as in the point #1.
  3. Ensure that an index exists on the serial_number column of the kickscooter_states_190614 table. Details about single-column / multi-column index remains the same as in the point #1.

Now, after ensuring these indexes, try your original query. Most likely, optimizer should be able to optimize the Join Order. Besides, the above query, you can enforce join order by using STRAIGHT_JOIN optimizer hint. So, try the following query as well, and benchmark between the two of them:

select
  r.user_id,
  k.id as kickscooter_id,
  st_astext(k.location) as location,
  k.created_at,
  k.serial_number,
  k_st.serial_number as states_serial_number,
  st_astext(k_st.gps) as gps_location,
  k_st.gps_updated_at,
  r.start_time,
  r.end_time
from kickscooters k
straight_join rents r
  on k.id= r.kickscooter_id
straight_join kickscooter_states_190614 k_st
  on k.serial_number = k_st.serial_number
order by r.rent_date
limit 999;

Upvotes: 1

Related Questions