As we know we are in the age of BIG DATA and slowly and steadily as the data gets bigger DWH environments will slowly find its way in cloud-based scale out systems.
In a typical DWH environment having very HUGE fact tables sometimes we want to keep several layers of hierarchy in the same table rather than recalculating the roll-ups each time .Here the usage of centipede fact tables are very essential which can store a number of hierararchy
within the same FACT table.In typical RDMS the method osf drawing data from table consists of 2 below mentioned methods as they are ROW oriented.
1.Selection
2.Projection
Due to the above mentioned property a databse like ORACLE or DB2 tends to select the entire (SELECTION) then it tends to select the REQUIRED COLUMNS.But in DWH environment this degrades the performance especially while we tend to use CENTIPEDE FACT tables.
As in particular SLICING report we tend to see a e few column esepecially in rolled-up queries.So here if ORACLE can come up with a concept of COLUMN-ORIENTATION or COLUMN-FAMILY based approach then we can have the luxury to follow the Method of PROJECTION wherein the column family or COLUMN group is taken first then the required number of rows taken according to the predicates.This will add a lot of performance to the rolled-up Queries and Performance enhancements of CENTIPEDE FACT table or Fast delivery of hierarchial reports.
The main theme of the thsi IDEA is that while we create a table in ORACLE by
CREATE TABLE(COL DATA_TYPE) command ,
a new clause can be introduced like ROW-RIENTED/COLUMN_ORIENTED can be introduced which will align the table according To ROW_FORMAT like normal ORACLE table or COLUMN-FORMATTED like HBASE or CASSANDRA according to appropriate situation @user9319583