CODEWITHSUNDEEP
mysqlindexingbigdata
blacksun
blacksun

Reputation: 733

Optimizing a very huge mysql table (query or mysql)

I have a mysql DB with 50 GB data and 200M record in a table. I am running following query and it takes 350 second to complete:

SELECT  x_date, count(*) as totl
    FROM  `rec_ex_15`
    WHERE  x_date > '2014-12-01'
      and  typx = '2'
    group by  x_date
    order by  x_date desc

x_date and typx are indexed.

Here is the explain:

id     select_type table       type     possible_keys  key     key_len  ref    rows
1      SIMPLE      rec_ex_15   range    typx,x_date    x_date  3        NULL   15896931  Using where

Are there any way to get the result faster?

Upvotes: 0

Views: 88

Answers (1)

DRapp
DRapp

Reputation: 48179

As noted in the comment... your query is very simple. You would be best to have a covering/compound index on (typx, x_date )

The typx = '2' first, then the index can jump directly to the date criteria and getting the results.

You may (future, consider creating a separate aggregate table, such as counts per type and date, then get aggregates from that if you are dealing with 200m records.

Clarifying the index (typx, x_date)

Since you are looking for a specific "typx", you want that in the first position of the index and secondary is the x_date. Think of the index like a room of boxes. Each box holds only 1 instance of a "typx" value (1, 2, 3, etc)... Within the box for (in your case typx = '2'), they are then sorted by x_date. So, within the box for typx = 2, you can jump directly to the date in question, grab the records and be done.

If the index were based on just x_date (as one you had available), assume each box in a room contains a single date. Yes, you can jump directly to the date starting your list, but then you have to go into the box and look through all entries to pull out any typx = 2 records. Now, close the box for day 1 and go to the next box date and look for any typx = 2 and so forth.

Having an efficient index based on your criteria can significantly help queries.

Upvotes: 2

Related Questions