How to reduce scan count on table with a composite index?

Question

We have a table (T) with ~3 million rows and just two INT columns (ID1 and ID2), which together are set as a composite clustered key.

Early in a stored procedure we create a table variable (@A) which consists of a list of INTs.

The slow query is the following

SELECT T.ID1, T.ID2 
FROM T
INNER JOIN @A AS A 
ON A.ID = T.ID1

@A will have just a few hundred rows, and T contains a few million rows. The problem is that T gets a scan count of several hundred. I don't know how to make that go away.

I have tried to create another index on T with column ID1 and ID2 included, but it does not help (the execution planner shows that the new index is used).

What can be done to reduce the scan count on table T?

(We are using SQL Server 2014, Web Edition)

Answer 1

Try creating an included (covering index):

create index idx_t_id1 on t(id1) include id2

This will allow your query to find everything it needs in the index pages and will not have to search in the main table. By the way is there a clustered index on table t?

Answer 2

You could try phrasing the query like this:

select a.id as id1,
       (select t.id2
        from t
        where t.id1 = a.id
       ) as id2
from @a a;

This should scan @a and use the index for the lookup.

Two notes:

• If there can be multiple matches, use cross apply instead.
• If there could be no matches and you want to filter the rows, use a subquery or CTE.

Answer 3

You should try temp table (#table) instead of table var (@table). SQL Server Always estimate @table as 1 row table and this could impact you query performance a lot. Another benefit to use #table is that you can create index on that;)