Joining two queries causes big performance hit

Question

There are two queries: one with a sub query and one without
although they both return the same data, one uses a full table scan of the indexed 'trade_index' table
and thus is much slower
any Ideas why it does this?:
SQL> select count(*) from trade_index ti, counterparty_xref cx
2 where ti.CPY_SID = cx.CPY_SID and ti.TRADE_OR_ARCHIVE = 'A'
3 and cx.XREF_VALUE in (select ml.VALUE from v_mis_lists ml where ml.LIST_NAME = 'Deutsche Bank DB')
4 and cx.XREF_TYPE = 'Client RXM'
5 and ti.START_DATE < to_date('01/01/2004','DD/MM/YYYY')
6 and (ti.END_DATE >= to_date('01/01/2003','DD/MM/YYYY') or ti.END_DATE = '01-Jan-1900') ;
COUNT(*)
6195
Elapsed: 00:00:15.08
Execution Plan
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2572 Card=1 Bytes=57)
1 0 SORT (AGGREGATE)
2 1 HASH JOIN (Cost=2572 Card=26581 Bytes=1515117)
3 2 HASH JOIN (Cost=460 Card=5619 Bytes=219141)
4 3 VIEW OF 'VW_NSO_1' (Cost=8 Card=16 Bytes=272)
5 4 SORT (UNIQUE) (Cost=8 Card=16 Bytes=960)
6 5 HASH JOIN (Cost=4 Card=16 Bytes=960)
7 6 TABLE ACCESS (BY INDEX ROWID) OF 'MIS_LISTS' (Cost=1 Card=2 Bytes=60)
8 7 INDEX (RANGE SCAN) OF 'MISL_NAT_UK' (UNIQUE) (Cost=2 Card=2)
9 6 TABLE ACCESS (FULL) OF 'MIS_LIST_VALUES' (Cost=2 Card=817 Bytes=24510)
10 3 TABLE ACCESS (FULL) OF 'COUNTERPARTY_XREF' (Cost=451 Card=35116 Bytes=772552)
11 2 TABLE ACCESS (FULL) OF 'TRADE_INDEX' (Cost=2056 Card=286928 Bytes=5164704)
SQL> select ml.VALUE from v_mis_lists ml where ml.LIST_NAME = 'Deutsche Bank DB';
VALUE
B49540
B60010
Elapsed: 00:00:00.00
Execution Plan
0 SELECT STATEMENT Optimizer=CHOOSE
1 0 NESTED LOOPS
2 1 TABLE ACCESS (BY INDEX ROWID) OF 'MIS_LISTS'
3 2 INDEX (RANGE SCAN) OF 'MISL_NAT_UK' (UNIQUE)
4 1 INDEX (RANGE SCAN) OF 'MISLV_PK' (UNIQUE)
SQL> select count(*) from trade_index ti, counterparty_xref cx
2 where ti.CPY_SID = cx.CPY_SID and ti.TRADE_OR_ARCHIVE = 'A'
3 and cx.XREF_VALUE in ('B49540','B60010')
4 and cx.XREF_TYPE = 'Client RXM'
5 and ti.START_DATE < to_date('01/01/2004','DD/MM/YYYY')
6 and (ti.END_DATE >= to_date('01/01/2003','DD/MM/YYYY') or ti.END_DATE = '01-Jan-1900') ;
COUNT(*)
6195
Elapsed: 00:00:02.06
Execution Plan
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=532 Card=1 Bytes=40)
1 0 SORT (AGGREGATE)
2 1 NESTED LOOPS (Cost=532 Card=41 Bytes=1640)
3 2 TABLE ACCESS (FULL) OF 'COUNTERPARTY_XREF' (Cost=451 Card=1 Bytes=22)
4 2 TABLE ACCESS (BY INDEX ROWID) OF 'TRADE_INDEX' (Cost=81 Card=286928 Bytes=5164704)
5 4 INDEX (RANGE SCAN) OF 'CPY_SID_IDX' (NON-UNIQUE) (Cost=2 Card=286928)
What I dont understand is that:
cx.XREF_VALUE in (select ml.VALUE from v_mis_lists ml where ml.LIST_NAME = 'Deutsche Bank DB')
and
cx.XREF_VALUE in ('B49540','B60010')
are essentially the same thing
(ie the subquery returns the values I have used)
but the optimizer uses a totally different (slower) method to execute the rest of the former query?
Many thanks
Ian

Answer 1

Hi,
Here is with join and not using the view: still have full table scan of large table trade_index
SQL> select count(*) from mis_lists ml, mis_list_values mlv, counterparty_xref cx, trade_index ti
2 where ti.CPY_SID = cx.CPY_SID and ti.TRADE_OR_ARCHIVE = 'A'
3 and cx.XREF_VALUE = mlv.VALUE
4 and mlv.MISL_SID = ml.MISL_SID
5 and ml.LIST_NAME = 'Deutsche Bank DB'
6 and cx.XREF_TYPE = 'Client RXM'
7 and ti.START_DATE < to_date('01/01/2004','DD/MM/YYYY')
8 and (ti.END_DATE >= to_date('01/01/2003','DD/MM/YYYY') or ti.END_DATE = '01-Jan-1900') ;
COUNT(*)
6195
Elapsed: 00:00:12.08
Execution Plan
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2602 Card=1 Bytes=10
0)
1 0 SORT (AGGREGATE)
2 1 HASH JOIN (Cost=2602 Card=26581 Bytes=2658100)
3 2 HASH JOIN (Cost=456 Card=5619 Bytes=460758)
4 3 HASH JOIN (Cost=4 Card=16 Bytes=960)
5 4 TABLE ACCESS (BY INDEX ROWID) OF 'MIS_LISTS' (Cost1 Card=2 Bytes=60)
6 5 INDEX (RANGE SCAN) OF 'MISL_NAT_UK' (UNIQUE) (Cost=2 Card=2)
7 4 TABLE ACCESS (FULL) OF 'MIS_LIST_VALUES' (Cost=2 Card=817 Bytes=24510)
8 3 TABLE ACCESS (FULL) OF 'COUNTERPARTY_XREF' (Cost=451Card=35116 Bytes=772552)
9 2 TABLE ACCESS (FULL) OF 'TRADE_INDEX' (Cost=2056 Card=286928 Bytes=5164704)
Statistics
0 recursive calls
47 db block gets
16519 consistent gets
4981 physical reads
0 redo size
205 bytes sent via SQL*Net to client
622 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
Thanks for you help,
Ian