AFFIC: A Foundation for Index Comparisons

Abstract

Comparing indexes (UB-Tree, R*-Tree, etc.) on a structural and theoretical level provides upper bounds for their worst-case performance, but it does not reflect the performance in real world applications. Measuring just query executions times is not fair either, but differences may be caused just by bad implementations. Thus, a fair comparison has to take into account the algorithms, their implementation, real world data, data management and queries and I/O behavior as well as run time.

Therefore, we need to take care that the implementations of indexes rely on a common foundation and differ just where there are conceptual differences. Furthermore, the foundation should be simple and flexible in order to allow adding new indexes easily while providing full control to all relevant parts. In our demo we present a small and flexible C++ library fulfilling these requirements. We expect to release this library in 2004 and thus provide a UB-Tree implementation to the database community useful for further research.