In partial fulfillment of the Requirements for the Degree of
Doctorate of Philosophy
will defend his dissertation proposal
Efficient OLAP Inside a Database Management System
On-Line Analytical Processing (OLAP) is a set of exploratory database techniques that allows the user to efficiently retrieve specific aggregations. In general, OLAP users analyze subsets of a dataset by grouping the data at different levels of detail in order to view the original dataset in different ways. The original intent of OLAP was to give users the ability to generate different types of reports, but these techniques are also being used to analyze datasets at different granularities. Currently OLAP is performed in either an external OLAP server or using external applications. The disadvantages of both methods are that they require the purchase of additional software other than the DBMS, the dataset must be exported in some fashion before being analyzed, and that they often require additional storage space because of the need to copy the datasets. Additionally, as the size of datasets continue to grow, we believe that it will become increasing infeasible to export and analyze the dataset outside the DBMS. To resolve this issue, we propose an alternative to these two methods that involves the execution of OLAP queries entirely within the DBMS using a combination of standard SQL, User Defined Functions, and Stored Procedures. This approach requires the creation of data structures that can either hold entire OLAP cubes in main memory or can transform the original dataset into a format that is more efficient for query processing. In our proposal, we will present the work that we have performed, some preliminary results, and a direction for our future research.