Help : Parallel query Syntax

Question

We have this huge table called CIFINV, with about 10 million records.
I need to move a bunch of data to another table like below
INSERT INTO CIF_VN_TEST( COL1, COL2 ) SELECT REPLACE(DSLTN,'-',''), STRIP_LAST_TOKN( KEY_VAL )
FROM CIFINV A WHERE A.TRACKINGNUMBER = (SELECT MAX(TRACKINGNUMBER)
FROM CIFINV WHERE DSLTN = A.DSLTN
AND SERVICEREQSUBTYPE IN ('INITIATE', 'MODIFY'))
AND UPPER(CLLI) IN ( select DISTINCT(UPPER(WIRE_CENTER)) FROM REG_SCHEDULE_LOOKUP b where TO_CHAR(b.MODE_START_DT,'MM-DD-YYYY')=TO_CHAR( sysdate, 'MM-DD-YYYY' )
Basically I need to select the rows
- with the maximum TRACKINGNUMBER for the given DSLTN
- Whose SERVICEREQSUBTYPE is either 'INITIATE' or 'MODIFY')
- Whose UPPER(CLLI ) exists in another table designated for processing for today
Im trying to parallelize this SQL.
Is this possible at all.
Please help

Answer 1

> We have this huge table called CIFINV, with about 10 million records.
Not really large, never mind huge. Billion row tables are considered large in terms of databases today.
And you seem to be under the incorrect impression that parallel processing magically solves problems. It does not. And only applies in specific conditions where Oracle is able to parallelise the SQL into multiple processes.
Also, what is missing is an execution plan. We have no idea how Oracle decides to execute your query and what part of the SQL is the expensive part.
> AND UPPER(CLLI) IN
This applies a function to a column to use for comparison purposes. Oracle cannot use an index that may exist on this column, as the indexed value is not the same as the uppercase value.
This likely forces Oracle to use scan the actual table itself. It is far better to not apply functions to the column values in predicates. Rather use those columns in their native data types and without any kind of function/transformation. This then enables Oracle to use any indexes that may exist on that column.
The same applies to apply functions to you MODE_START_DT column in the sub-select. Don't. Use a function on that column and you loose the ability to use the index.
Instead use the data predicate as follows:
WHERE b.mode_start_dt between TRUNC(SYSDATE) and TRUNC(SYSDATE)+1-(1/60/60/24)
Consider not use UPPER on the the CLLI column. This will break an index on CLLI. Alternatively, create a function index, e.g.
create index <indexname> on <table> ( <function>( <column> ) )
Also consider what indexes to use - bitmap, b+tree, function, etc.
I suggest that you look for the thread "Why my qeury is slow" (on this forum - use the search function), and follow the instructions in it. It tells you what data you need to supply the forum in order for us to provide you with a meaningful answer regarding your SQL performance problem.