如何加快 sql 查询？

Question

I have an SQL query as follows:

SELECT p.Id1,p.Id2,p.Id3 
FROM dataset1 p
WHERE p.Id2 IN (
    SELECT r.Id4 
    FROM dataset1 r 
    WHERE r.Id5=125 AND r.Id6>=100000000000000 AND r.Id6<1000000000000000
) 
ORDER BY p.Id1 DESC, p.Id2 DESC

However there appears to be huge amounts of data with Id6 in this range and thus, it takes a quite long time to compute. But I only have one hour to compute the query. I thus, am wondering if someone could help me to improve the performance of this query.

Thanks.

Answer 1

Since the filtering seems to be done on r , arrange for it to be looked at first:

SELECT  p.Id1, p.Id2, p.Id3
    FROM  ( SELECT id4
       FROM dataset1 AS r
       WHERE  r.id5 = 125
         AND  r.Id6 >= 100000000000000
         AND  r.Id6 <  100000000000000 ) AS x
    JOIN dataset1 AS p  ON p.id2 = x.id4
    ORDER BY  p.Id1 DESC, p.Id2 DESC;

For that, these indexes should be beneficial:

INDEX(id5, id6, id4)   -- covering
INDEX(id2, id1, id3)   -- covering

You have a "range" test on id6 , yet the range is empty. I assume that was a mistake. Please don't simplify a query too much; we might give you advice that does not apply. I am assuming that the range is really a range.

Answer 2

IN tend to optimize poorly when the subquery returns a lot of data. You can try using EXISTS instead:

SELECT p.Id1, p.Id2, p.Id3 
FROM dataset1 p
WHERE EXISTS (
    SELECT 1
    FROM dataset1 r 
    WHERE 
        r.Id4 = p.Id2
        AND r.Id5 = 125 
        AND r.Id6 >= 100000000000000 
        AND r.Id6 <  100000000000000
) 
ORDER BY p.Id1 DESC, p.Id2 DESC

Then, consider a multi-column index on (Id4, Id5, Id6) to speed up the subquery. The idea is to put the more restrictive criteria first - so obviously you want Id6 last, but you might want to try inverting the first two columns to see if any combination performs better than the other.

Side note: both the lower and upper bound for Id6 have the same value in your query. I take this as a typo (otherwise your query would always return no row).

Answer 3

To improve performance don't use an inner query. You can get you desired result by using an inner join too:

SELECT 
    p.Id1, p.Id2, p.Id3 
FROM 
    dataset1 p 
INNER JOIN 
    dataset1 r ON p.Id2 = r.Id4 
               AND r.Id5 = 125 
               AND r.Id6 >= 100000000000000 
               AND r.Id6 < 100000000000000
ORDER BY 
    p.Id1 DESC, p.Id2 DESC