On Modeling and Predicting Query Behavior in OLAP Systems

Interactive multidimensional data analysis tools (mostly OLAP systems) are the predominant frontend tools for end users in data warehouse environments. Thus, the design of these systems is an important part of the data warehouse design itself. This paper contributes to the important design step by discussing the modeling of user query behavior and its benefits. We present a mathematical model and a graphical notation for capturing knowledge about typical multidimensional interaction patterns in OLAP systems, taking into account the session oriented, interactive and navigational nature of the user query behavior. To exemplarily show that the modeling of user behavior is not only beneficial during the logical and physical design phase, we also present an architecture to speed up OLAP systems at runtime by using speculative execution techniques based on a prediction of the user query behavior.