A nested loop join is performed by doing a scan over the left subtree and for each row in it performing a full scan of the right subtree.

This is the default join algorithm, which can be used for any join. However, it is usually less efficient than the other methods. Usually, either an existing index or a dynamic index, used in an ANL join, will cost much less. Occasionally, when subtree cardinalities are very low, possibly because of index bracketing, nested loop will be the method with the least cost.