HELp to reslove this ETL problem.

Based on the topics we have learned on ETL, come up with a best optimal approach to load following historical data onto Data Warehouse Fact tables?

Scenario: As part of initial setup of a DW, We have to extract 3 years of data (history) from a source system and load onto Data warehouse fact table.

On average each year has 1.2 Crore rows with each row size being 1024 bytes.

2 months of data processing has following statistics:

-          4 hours to  extract from source system

-          30 min for data transfer from source to a system to target system. (file copy)

Assumption:

·        You have the latest hardware with multiple processors, parallelism (4) on both source (Mainframe system) and the Data Warehouse (UNIX) systems.