Building the Data Warehouse

Скачать в pdf «Building the Data Warehouse»


By creating two levels of granularity in the data warehouse, the DSS architect has killed two birds with one stone—most DSS processing is performed against

Figure 2.17 Dual levels of granularity allow you to process the majority of requests efficiently and answer any question that can be answered.


the lightly summarized data, where the data is compact and efficiently accessed. When some greater level of detail is required—5 percent of the time or less—there is the true archival level of data. It is expensive, cumbersome, and complex to access data at the true archival level of granularity, but if there it is there when necessary.


If a pattern of searching the true archival level of data develops over time, the designer may want to create some new fields of data at the lightly summarized level, so that most of the processing can occur there.


Because of the costs, efficiencies, ease of access, and ability to answer any query that can be answered, the dual level of data is the best architectural choice for the detailed level of the data warehouse for most shops. A single


level of data should only be attempted when a shop has a relatively small amount of data in the data warehouse environment.

Exploration and Data Mining


The granular data found in the data warehouse supports more than data marts. It also supports the processes of exploration and data mining. Exploration and data mining take masses of detailed, historical data and examine it for previously unknown patterns of business activity.

Скачать в pdf «Building the Data Warehouse»