What combination of partitioning Method is better?

Question

Hi everybody,
I have the following problem:
I have a big table (more than 1000 000 000 lines).
Create table sales (
Id_dept varchar(1),
sales_day date,
day varchar(2),
Qte number(5)
I would partition this table. I have only 9 departments and I load data day by day. What combination off partitioning is better?
Solution 1) RANGE partitioning by date and LIST sub partitioning by department.
Or
Solution 2) RANGE Partitioning by department (id_detp) and HASH sub partitioning by date.
The first solution gives me administration tasks facility (add and drop partitions will be easy: day by day).
The second solution seems to me better for partition pruning.
95% of requests concern only one department like the following request:
Select * from sales where id_dept= :x
and sales_day in (…), and …
In solution 2 I filter immediately 8 other department partitions.
But as I have only 9 departments, I wonder if the range partitioning is ok for department. In other word is the range partitioning algorithm is goes well to use only one value in place of a range of values?
Partition by RANGE (id_dept)
Subpartition by HASH(sales_day) subpartitions 90
(Partition part1 values less than (2),
Partition part2 values less than (3),
Partition part3 values less than (4),
Partition part9 values less than (9));
Could you give me on advice please
Regards
Azizollah

Answer 1

Solution 2 is out of the question as you've stated it. First, you would typically range partition by a value such as a date and no need for a Hash when you have fields which lend themselves to logical segregation (sales_day and Id_dept). A hash is best for tables in which you don't have fields that provide any sort of logical separation.
If you are partitioning, it is basically to make managing your tables easier (e.g.. partition pruning). Therefore, pruning by date and not by dept makes more sense. Without knowing exactly what you want the partitioning for or how you are using your tables, can't say do this or that. But given your table, I'd take a look at doing a composite of range (date) as the top-level scheme with list (dept) as the sub-partition.
For documentation, refer to:
http://download-east.oracle.com/docs/cd/B19306_01/server.102/b14223/parpart.htm#g1020112