Does a foreign key automatically create an index?

417
votes

I've been told that if I foreign key two tables, that SQL Server will create something akin to an index in the child table. I have a hard time believing this to be true, but can't find much out there related specifically to this.

My real reason for asking this is because we're experiencing some very slow response time in a delete statement against a table that has probably 15 related tables. I've asked our database guy and he says that if there is a foreign key on the fields, then it acts like an index. What is your experience with this? Should I add indexes on all foreign key fields or are they just unnecessary overhead?

11
sql-server
I have the same understanding as your DB guy - that FKs do in fact create an index. – Vinnie
No - a FK does NOT automatically create an index. It makes sense to create one - but it is NOT done automatically by SQL Server. – marc_s
not silly to ask this at all ! – marc_s
If you are getting slow deletes and the table you are deleting from referenced by other tables, you will probably get a performance boost by indexing the foreign keys in the other tables. This is because when SQL is deleting a row, it needs to check referential integrity on the row. To do this, it obviously needs to check that no other rows exist referencing the row you are deleting. – Noel Kennedy
I'd say a database guy who didn't know this is must be in serious need of training. Database people are responsible for performance, it is their job to know this sort of thing. This suggests gross incompetence. – HLGEM

11 Answers

373
votes

A foreign key is a constraint, a relationship between two tables - that has nothing to do with an index per se.

But it is a known fact that it makes a lot of sense to index all the columns that are part of any foreign key relationship, because through a FK-relationship, you'll often need to lookup a relating table and extract certain rows based on a single value or a range of values.

So it makes good sense to index any columns involved in a FK, but a FK per se is not an index.

Check out Kimberly Tripp's excellent article "When did SQL Server stop putting indexes on Foreign Key columns?".

45
votes

Wow, the answers are all over the map. So the Documentation says:

A FOREIGN KEY constraint is a candidate for an index because:

  • Changes to PRIMARY KEY constraints are checked with FOREIGN KEY constraints in related tables.

  • Foreign key columns are often used in join criteria when the data from related tables is combined in queries by matching the column(s) in the FOREIGN KEY constraint of one table with the primary or unique key column(s) in the other table. An index allows Microsoft® SQL Server™ 2000 to find related data in the foreign key table quickly. However, creating this index is not a requirement. Data from two related tables can be combined even if no PRIMARY KEY or FOREIGN KEY constraints are defined between the tables, but a foreign key relationship between two tables indicates that the two tables have been optimized to be combined in a query that uses the keys as its criteria.

So it seems pretty clear (although the documentation is a bit muddled) that it does not in fact create an index.

21
votes

No, there is no implicit index on foreign key fields, otherwise why would Microsoft say "Creating an index on a foreign key is often useful". Your colleague may be confusing the foreign key field in the referring table with the primary key in the referred-to table - primary keys do create an implicit index.

11
votes

Foreign keys do not create indexes. Only alternate key constraints(UNIQUE) and primary key constraints create indexes. This is true in Oracle and SQL Server.

7
votes

SQL Server autocreates indices for Primary Keys, but not for Foreign Keys. Create the index for the Foreign Keys. It's probably worth the overhead.

6
votes

Say you have a big table called orders, and a small table called customers. There is a foreign key from an order to a customer. Now if you delete a customer, Sql Server must check that there are no orphan orders; if there are, it raises an error.

To check if there are any orders, Sql Server has to search the big orders table. Now if there is an index, the search will be fast; if there is not, the search will be slow.

So in this case, the slow delete could be explained by the absence of an index. Especially if Sql Server would have to search 15 big tables without an index.

P.S. If the foreign key has ON DELETE CASCADE, Sql Server still has to search the order table, but then to remove any orders that reference the deleted customer.

4
votes

Strictly speaking, foreign keys have absolutely nothing to do with indexes, yes. But, as the speakers above me pointed out, it makes sense to create one to speed up the FK-lookups. In fact, in MySQL, if you don't specify an index in your FK declaration, the engine (InnoDB) creates it for you automatically.

4
votes

In PostgeSql you can check for indexes yourself if you hit \d tablename

You will see that btree indexes have been automatically created on columns with primary key and unique constraints, but not on columns with foreign keys.

I think that answers your question at least for postgres.

3
votes

Not to my knowledge. A foreign key only adds a constraint that the value in the child key also be represented somewhere in the parent column. It's not telling the database that the child key also needs to be indexed, only constrained.

2
votes

I notice that Entity Framework 6.1 pointed at MSSQL does automatically add indexes on foreign keys.

1
votes

It depends. On MySQL an index is created if you don't create it on your own:

MySQL requires that foreign key columns be indexed; if you create a table with a foreign key constraint but no index on a given column, an index is created.

Source: https://dev.mysql.com/doc/refman/8.0/en/constraint-foreign-key.html

The same for MySQL 5.6 eh.