Building the Data Warehouse

Скачать в pdf «Building the Data Warehouse»

The data model exists at three levels—high level, midlevel, and low level. The data model is the key to being able to build the data warehouse in iterations. The entities found in the high-level model relate to the major subject areas of the corporation. The low-level model relates to the physical database design of the data warehouse.

At the lowest level of database design, slight denormalization can occur when the entire organization looks at the data in the same way. Some techniques for slight denormalization of data include creating arrays of data, creating redundant data judiciously, and making creative indexes.

The basic structure of the data warehouse record is one that contains a timestamp, a key, direct data, and secondary data. All data warehouse database designs—in one form or the other—follow this simple pattern.

Reference tables need to be placed in the data warehouse and managed on a time-variant basis just like any other data. There are many approaches to the inclusion and design of reference data in the data warehouse.

Data is loaded into the data warehouse under what can be termed a “wrinkle of time.” This means that as soon as an activity occurs in the operational environment, that data is not immediately rushed to the data warehouse. Instead, data

that has been newly updated in the operational environment is allowed to stand in the operational environment for up to 24 hours before being moved to the data warehouse.

The transformation that occurs as data moves from the operational environment to the data warehouse environment is complex. There is a change in DBMS, a change in operating systems, a change in hardware architecture, a change in semantics, a change in coding, and so forth. Many, many considerations are involved in the movement of data from the operational environment to the data warehouse environment.

Скачать в pdf «Building the Data Warehouse»