Relational vs OLAP in the Dimensional World

Presentation Summary

Steve Grantham, Boise State University

Dimensional data models can be implemented within either a relational database or an OLAP cube architecture; each has its own strengths and weaknesses. To a first approximation, the relational layer is sort of like a well used 4WD pickup truck; you can get almost anywhere with it, and carry almost anything, but it may not be the fastest way and the ride may be a bit bumpy. The OLAP layer is more like a sports car: on the right roads it’s faster and more fun, but it takes a bit of maintenance to make it run right, and there are some places you just can’t take it and some things you can’t do with it. In this presentation I will illustrate this analogy with some specific examples of various types of reporting needs and which layer supports each one best. This will be based on Boise State’s implementation of the iStrategy student data warehouse, but I believe the principles illustrated are fairly general.