Building the Data Warehouse

Скачать в pdf «Building the Data Warehouse»

Called a dual level of granularity, the design shown in Figure 2.15—a phone company—fits the needs of most shops. There is a tremendous amount of detail at the operational level. Most of this detail is needed for the billing systems. Up to 30 days of detail is stored in the operational level.

The data warehouse in this example contains two types of data-lightly summarized data and “true archival” detail data. The data in the data warehouse can go back 10 years. The data that emanates from the data warehouse is “district” data that flows to the different districts of the telephone company. Each district

high level of detail

low level of detail

level of detail—answer

flexibility—small enough

any question

volumes to be able to be


large volumes of data

small volumes

a telephone company

dual levels of granularity

Figure 2.15

The volume of data is such that most organizations need to have two levels of granularity in the data warehouse.

then analyzes its data independently from other districts. Much heuristic analytical processing occurs at the individual level.

A light summarization data is detailed data that has been summarized only to a very small extent. For example, phone call information may be summarized by the hour. Or bank checking information may be summarized by the day. Figure 2.16 shows such a summarization.

As data is passed from the operational, 30-day store of data, it is summarized, by customer, into fields that are likely to be used for DSS analysis. The record for J Jones shows the number of calls made per month, the average length of each call, the number of long-distance calls made, the number of operator-assisted calls, and so forth.

Скачать в pdf «Building the Data Warehouse»