Building the Data Warehouse

Скачать в pdf «Building the Data Warehouse»


means nothing because the processing that operates against the living sample does not require every record in the data warehouse to be in the living sample.


The greatest asset of a living sample database is that it is very efficient to access. Because its size is a fraction of the larger database from which it was derived, it is correspondingly much more efficient to access and analyze.


Put another way, suppose an analyst takes 24 hours to scan and analyze a large database. It may take as little as 10 minutes to scan and analyze a living sample database. In doing heuristic analysis, the turnaround time is crucial to the analysis that can be done. In heuristic analysis, the analyst runs a program, studies the results, reformulates the program, and runs it again. If it takes 24 hours to execute the program, the process of analysis and reformulation is greatly impaired (not to mention the resources required to do the reformulation).


With a living sample database small enough to be scanned in 10 minutes, the analyst can go through the iterative process very quickly. In short, the productivity of the DSS analyst depends on the speed of turning around the analysis being done.


One argument claims that doing statistical analysis yields incorrect answers. For example, an analyst may run against a large file of 25 million records to determine that 56.7 percent of the drivers on the road are men. Using a living sample database, the analyst uses 25,000 records to determine that 55.9 percent of the drivers on the road are men. One analysis has required vastly more resources than the other, yet the difference between the calculations is very small. Undoubtedly, the analysis against the large database was more accurate, but the cost of that accuracy is exorbitant, especially in the face of heuristic processing, where iterations of processing are the norm.

Скачать в pdf «Building the Data Warehouse»