今天在ORACLE8.17上一个做一个统计SQL，开始用

Select /*+ ordered  parallel 8 index (a index_a) */
     a.msisdn  From dbuser.table_b   b,
               dbuser.table_a a
       Where  a.serv_id=b.serv_id And A.BILLING_CYCLE_ID<'200411' And a.billing_cycle_id>'200406'
执行计划如下

SELECT STATEMENT, GOAL = RULE   Cost=1221 Cardinality=120 Bytes=3720
 HASH JOIN   Cost=1221 Cardinality=120 Bytes=3720
  TABLE ACCESS FULL Object wner=ACCOUNT Object name=table_b Cost=391 Cardinality=2000 Bytes=16000
  TABLE ACCESS BY INDEX ROWID Object wner=ACCOUNT Object name=table_a Cost=826 Cardinality=6 Bytes=138
   INDEX FULL SCAN Object wner=SALES Object name=index_a Cost=26 Cardinality=6 

非常奇怪，居然使用的是INDEX FULL SCAN 

然后尝试使用no_index ,排除可能使用索引

Select /*+ ordered  parallel 8 no_index (a index_b) */
     a.msisdn  From dbuser.table_b   b,
               dbuser.table_a a
       Where  a.serv_id=b.serv_id And A.BILLING_CYCLE_ID<'200411' And a.billing_cycle_id>'200406'
 
 结果居然是我想要的，执行计划如下

SELECT STATEMENT, GOAL = RULE   Cost=4391 Cardinality=120 Bytes=3720
 NESTED LOOPS   Cost=4391 Cardinality=120 Bytes=3720
  TABLE ACCESS FULL Object wner=ACCOUNT Object name=table_b Cost=391 Cardinality=2000 Bytes=16000
  TABLE ACCESS BY INDEX ROWID Object wner=ACCOUNT Object name=table_a Cost=2 Cardinality=6 Bytes=138
   INDEX RANGE SCAN Object wner=SALES Object name=index_a Cost=1 Cardinality=6