Parellelism vs multiple queries using scheduler

Question

Hi,
I am trying to run a query that is going to insert the data in a table (TEMP_RESULT).
At very high level lets just say that I have compare each row from "t_customer" with each row in "t_region". The result is inserted into a table.
I am HASH PARTITIONING "t_customer" (32 partitions) on 'Customer_ID' and "t_region" on 'REGION_ID' (16 partitions).
I am trying to get the best timing possible. The question is should I just let oracle use PARALLELISM and run this in one chunk or should I loop?
My current approcah is I loop through each 'Customer_ID' and compare it against all 'REGION_ID'. Assume nothing else is running on the Database, then what otion would be the best:
1) Let it just run through 1 process?
2) Would it be bettter if I start a second process for another 'CUSTOMER_ID' and have a few of these running simultaneously?
3) Another option I am trying to consider is to run 4 processes each process will handle a partition of the REGION table. I just need to somehow identify which 'REGION_ID' would fall in what partition and pass the values such that I have my where clasue something like "WHERE region_id between :min_region_id and :max_region_id"
with t_customer as
The customer table.  For the sake of simplicity I only have 5 codes per customer.
My real values are 7385 customers with 1269 codes.
Total number of rows is 7385*1269 = 9,371,565
SELECT 1 as customer_id, 201 as internal_code, 1 as fee from dual union all
SELECT 1 as customer_id, 202 as internal_code, 3 as fee from dual union all
SELECT 1 as customer_id, 203 as internal_code, 5 as fee from dual union all
SELECT 1 as customer_id, 204 as internal_code, 7 as fee from dual union all
SELECT 1 as customer_id, 205 as internal_code, 11 as fee from dual union all
SELECT 2 as customer_id, 201 as internal_code, 11 as fee from dual union all
SELECT 2 as customer_id, 202 as internal_code, 7 as fee from dual union all
SELECT 2 as customer_id, 203 as internal_code, 5 as fee from dual union all
SELECT 2 as customer_id, 204 as internal_code, 3 as fee from dual union all
SELECT 2 as customer_id, 205 as internal_code, 1 as fee from dual
, t_region as
The customer table.  For the sake of simplicity I only have 5 codes per region.
My real values are 3497 regions with 1269 codes.
Total number of rows is 3497*1269 = 4,437,693
SELECT 95 as region_id, 201 as internal_code, 0 as fee from dual union all
SELECT 95 as region_id, 202 as internal_code, 2 as fee from dual union all
SELECT 95 as region_id, 203 as internal_code, 4 as fee from dual union all
SELECT 95 as region_id, 204 as internal_code, 6 as fee from dual union all
SELECT 95 as region_id, 205 as internal_code, 8 as fee from dual union all
SELECT 96 as region_id, 201 as internal_code, 8 as fee from dual union all
SELECT 96 as region_id, 202 as internal_code, 6 as fee from dual union all
SELECT 96 as region_id, 203 as internal_code, 4 as fee from dual union all
SELECT 96 as region_id, 204 as internal_code, 2 as fee from dual union all
SELECT 96 as region_id, 205 as internal_code, 0 as fee from dual
The claculation
Compare each cutomer with each region for each code
Let's keep that simple too.
Row Count is Count(Distinct Customer ID) * Count(Distinct Region ID) * Count(Distinct Internal Code)
Sample: 2 * 2 * 5 = 20
Real:   7385 * 3497 * 1269 = 32,772,362,805
SELECT
         c.customer_id
       , r.region_id
       , c.internal_code
       , c.fee AS customer_fee
       , r.fee AS region_fee
       , c.fee - r.fee AS diff
    FROM t_customer c INNER JOIN t_region r ON c.internal_code = r.internal_codeWell there it is a sample that shows my problem, at a very basic level. The calculation itself is not very complicated, but what is stalling me is the size of the data.
Thanks.

Answer 1

The calculation is not as simple as it was stated here. I do need to store it, becuase we want to keep track of changes to the calculation.
Here is what the calculation is really like.
You were right, just the the base join the PARTITION-WISE join worked great. Howevere, I have to also group the data with another set. That's where the timing is really killing me.
The "internal_code" table is only about 12,000 records, do you think partitioning that table would be of any help.
Any suggestions would be helpful.
with t_customer as
The customer table.  For the sake of simplicity I only have 5 codes per customer.
My real values are 7385 customers with 1269 codes.
Total number of rows is 7385*1269 = 9,371,565
SELECT 1 as customer_id, 201 as internal_code, 1 as fee, 5 as base_fee, 10 as pct, 5 as default_pct from dual union all
SELECT 1 as customer_id, 202 as internal_code, 0 as fee, 5 as base_fee, 10 as pct, 5 as default_pct from dual union all
SELECT 1 as customer_id, 203 as internal_code, 5 as fee, 5 as base_fee, 10 as pct, 5 as default_pct from dual union all
SELECT 1 as customer_id, 204 as internal_code, 7 as fee, 5 as base_fee, 10 as pct, 5 as default_pct from dual union all
SELECT 1 as customer_id, 205 as internal_code, 11 as fee,5 as base_fee, 10 as pct, 5 as default_pct  from dual union all
SELECT 2 as customer_id, 201 as internal_code, 11 as fee,5 as base_fee, 10 as pct, 5 as default_pct  from dual union all
SELECT 2 as customer_id, 202 as internal_code, 0 as fee, 5 as base_fee, 10 as pct, 5 as default_pct  from dual union all
SELECT 2 as customer_id, 203 as internal_code, 5 as fee, 5 as base_fee, 10 as pct, 5 as default_pct  from dual union all
SELECT 2 as customer_id, 204 as internal_code, 3 as fee, 5 as base_fee, 10 as pct, 5 as default_pct  from dual union all
SELECT 2 as customer_id, 205 as internal_code, 1 as fee, 5 as base_fee, 10 as pct, 5 as default_pct  from dual
, t_region as
The customer table.  For the sake of simplicity I only have 5 codes per region.
My real values are 3497 regions with 1269 codes.
Total number of rows is 3497*1269 = 4,437,693
SELECT 95 as region_id, 201 as internal_code, 0 as fee, 5 as pct from dual union all
SELECT 95 as region_id, 202 as internal_code, 2 as fee, 5 as pct from dual union all
SELECT 95 as region_id, 203 as internal_code, 4 as fee, 5 as pct from dual union all
SELECT 95 as region_id, 204 as internal_code, 6 as fee, 5 as pct from dual union all
SELECT 95 as region_id, 205 as internal_code, 8 as fee, 5 as pct from dual union all
SELECT 96 as region_id, 201 as internal_code, 8 as fee, 5 as pct from dual union all
SELECT 96 as region_id, 202 as internal_code, 6 as fee, 5 as pct from dual union all
SELECT 96 as region_id, 203 as internal_code, 4 as fee, 5 as pct from dual union all
SELECT 96 as region_id, 204 as internal_code, 2 as fee, 5 as pct from dual union all
SELECT 96 as region_id, 205 as internal_code, 0 as fee, 5 as pct from dual
There are about 1269 unique internal_codes and they belong to different groups.
The table size for internal code is 12,000 records.
, t_internal_code
as
SELECT 1 as grp_id, 201 as internal_code, .5 as rate FROM DUAL UNION ALL
SELECT 1 as grp_id, 202 as internal_code, .5 as rate FROM DUAL UNION ALL
SELECT 1 as grp_id, 203 as internal_code, .2 as rate FROM DUAL UNION ALL
SELECT 1 as grp_id, 204 as internal_code, .3 as rate FROM DUAL UNION ALL
SELECT 1 as grp_id, 205 as internal_code, .1 as rate FROM DUAL UNION ALL
SELECT 2 as grp_id, 201 as internal_code, .5 as rate FROM DUAL UNION ALL
SELECT 2 as grp_id, 202 as internal_code, .5 as rate FROM DUAL
   , t_base_calc AS 
The claculation
Compare each cutomer with each region for each code
Let's keep that simple too.
Row Count is Count(Distinct Customer ID) * Count(Distinct Region ID) * Count(Distinct Internal Code)
Sample: 2 * 2 * 5 = 20
Real:   7385 * 3497 * 1269 = 32,772,362,805
     (SELECT customer_id
           , region_id
           , internal_code
           , customer_adjusted_fee
           , region_adjusted_fee
           , customer_adjusted_fee - region_adjusted_fee AS base_calc
        FROM (SELECT c.customer_id
                   , r.region_id
                   , c.internal_code
                   , c.fee AS customer_fee
                   , r.fee AS region_fee
                   , c.pct AS customer_pct
                   , r.pct AS region_pct
                   , c.base_fee
                   , c.default_pct
                   , CASE c.fee
                        WHEN 0
                           THEN (CASE r.fee
                                    WHEN 0
                                       THEN (CASE c.pct
                                                WHEN 0
                                                   THEN c.default_pct
                                                ELSE c.pct
                                             END) * c.base_fee / 100
                                    ELSE r.fee
                                 END) * c.pct / 100
                        ELSE c.fee
                     END AS customer_adjusted_fee
                   , CASE r.fee
                        WHEN 0
                           THEN (CASE c.pct
                                    WHEN 0
                                       THEN c.default_pct
                                    ELSE c.pct
                                 END) * c.base_fee / 100
                        ELSE r.fee
                     END AS region_adjusted_fee
                FROM t_customer c INNER JOIN t_region r ON c.internal_code = r.internal_code
             ) z
       WHERE customer_adjusted_fee - region_adjusted_fee > 0)
   , t_calcs AS
     (SELECT   bc.customer_id
             , bc.region_id
             , ic.grp_id
             , SUM (ic.rate) AS calc_2
             , SUM (bc.base_calc * ic.rate) / SUM (ic.rate) AS calc_3
             , SUM (CASE
                       WHEN bc.base_calc > 50
                          THEN ic.rate
                       ELSE 0
                    END) AS calc_4
          FROM t_base_calc bc INNER JOIN t_internal_code ic ON bc.internal_code = ic.internal_code
      GROUP BY bc.customer_id
             , bc.region_id
             , ic.grp_id)
SELECT *
  FROM t_calcsMessage was edited by:
pshah2k