Data Warehousing for Not Yet Techies

Data warehousing is the process of separating and storing data in a central repository. Data warehousing was coined by W.H. Inmon and is typically stored on a mainframe server.

Data warehousing separates operational databasing such as what goes on in minute by minute business transactions from the analytical processing of data. When you are using a data warehouse, you don't want the information constantly changed by new sales, changes of addresses etc.

Data warehousing supports data mining and a Decision Support System (DSS), which is analytical processing. It is at the heart of DSS analytical processing.

Two components are always analyzed: the data being warehoused and the usage of that data.

• Subject-oriented
• Integrated
• Nonvolatile
• Time-Variant

It uses a much longer time frame that operational databases. Data warehousing has several design issues:

Granularity -- this is the level of detail supported by data mining. For instance, if you want to know which region of the country is the best market for your product, you need to know whether customers are in the North, South, West, Midwest etc -- but not their exact address. But if you are analyzing the effectiveness of a mailing list to sell your product to, you want to their zip code. Because, if your product is expensive people in poor zip codes can't afford it. If your product appeals to poor people, you don't want to send it to people in wealthy zip codes.

A living sample database is a sample of a much larger database. Since every query and report you run on your data costs money, it may be more cost-effective to query a statistically significant -sized living sample database.

Partitioning -- this is breaking the data up into separate physical units.

Data warehousing is necessary for large businesses with terabytes of data to store and use.

Next: FrontPage