oracle数据库use_hash,USE_HASH的执行计划,为什么要这么走呢?

SQL> select /*+USE_HASH (lf rf)*/lf.a al,lf.b bl,lf.c cl

2  from test_llf lf,test_rrf rf

3  where lf.a=rf.a(+)

4  union

5  select  /*+USE_HASH (lf rf)*/rf.a ar,rf.b br,rf.c cr

6  from test_llf lf,test_rrf rf

7  where rf.a=lf.a(+);

7 rows selected.

Execution Plan

----------------------------------------------------------

0      SELECT STATEMENT Optimizer=ALL_ROWS (Cost=11 Card=7 Bytes=36

1    0   SORT (UNIQUE) (Cost=11 Card=7 Bytes=364)

2    1     UNION-ALL

3    2       HASH JOIN (OUTER) (Cost=5 Card=4 Bytes=208)

4    3         TABLE ACCESS (FULL) OF 'TEST_LLF' (TABLE) (Cost=2 Ca

5    3         INDEX (FAST FULL SCAN) OF 'SYS_C005288' (INDEX (UNIQ

6    2       HASH JOIN (OUTER) (Cost=5 Card=3 Bytes=156)

7    6         TABLE ACCESS (FULL) OF 'TEST_RRF' (TABLE) (Cost=2 Ca

8    6         INDEX (FAST FULL SCAN) OF 'SYS_C005287' (INDEX (UNIQ

Statistics

----------------------------------------------------------

0  recursive calls

0  db block gets

12  consistent gets

0  physical reads

0  redo size

612  bytes sent via SQL*Net to client

512  bytes received via SQL*Net from client

2  SQL*Net roundtrips to/from client

1  sorts (memory)

0  sorts (disk)

7  rows processed

请问,为什么走hash join时候会发生?

TABLE ACCESS (FULL) OF 'TEST_LLF' (TABLE)

TABLE ACCESS (FULL) OF 'TEST_RRF'

有没有办法不要走TABLE ACCESS ,在两个表都有几百万行的数据情况下,如果改善?