Shared scans on main memory column stores
Gustavo Alonso, Donald Kossmann, Tudor-Ioan Salomie, Andreas U. Schmidt · Repository for Publications and Research Data (ETH Zurich) · 2012
Column stores and shared scans have been found to be effective techniques in order to improve performance for many workloads. Another recent hardware trend makes it possible to keep most data in main memory. This paper builds upon these trends and explores how to implement shared scans on column stores in main memory efficiently. In particular, this paper proposes new approaches to avoid unnecessary work and to best implement position lists in such a query processing archi tecture. Performance experiments with real workloads from the travel industry show the advantages of combining column stores and shared scans in main memory over traditional database architectures.