Abstract

Online Analytical Processing (OLAP) is a database paradigm that supports the
rich analysis of multi-dimensional data. While current OLAP tools are primarily
constructed as extensions to conventional relational databases, the unique modeling
and processing requirements of OLAP systems often make for a relatively awkward
fit with RDBM systems in general, and their embedded string-based query languages
in particular. In this thesis, we discuss the design, implementation, and evaluation
of a robust multi-dimensional OLAP server. In fact, we focus on several distinct but
related themes. To begin, we investigate the integration of an open source embedded
storage engine with our own OLAP-specific indexing and access methods. We then
present a comprehensive OLAP query algebra that ultimately allows developers to
create expressive OLAP queries in native client languages such as Java. By utilizing
a formal algebraic model, we are able to support an intuitive Object Oriented query
API, as well as a powerful query optimization and execution engine. The thesis
describes both the optimization methodology and the related algorithms for the
efficient execution of the associated query plans. The end result of our research is a
comprehensive OLAP DBMS prototype that clearly demonstrates new opportunities
for improving the accessibility, functionality, and performance of current OLAP database
management systems.