Computing Seminar

Bitmap Indices for Speeding Up End User Physics Analysis

Kurt Stockinger/IT

Abstract

Bitmap indices are index structures that efficiently support complex, multi-dimensional queries for datawarehouse applications. These data structures are also implemented in commercial database management systems such as Oracle, Sybase or Informics. All these implementations are optimised for typical business applications which are characterised by discrete attribute values. However, scientific data which is mostly characterised by non-discrete attribute values, cannot be handled efficiently by these kind of data structures.

In this seminar we give an overview of bitmap indices and demonstrate that this kind of access method shows good performance behaviour in multi-dimensional search spaces and significantly outperforms the sequential scan for typical end user physics analysis.

About the speaker: Kurt Stockinger graduated in computer science and business administration from the University of Vienna in 1999. He then did his Ph.D. in the CERN/IT/DB database group on access optimisation based on bitmap indices. He is a member of WP2 (Grid Data Management) of the EU DataGrid Project. Currently he works as a fellow in the database group on Grid query optimisation.