Exploiting the multi-append-only-trend property of historical data in data warehouses

Abstract

Data warehouses maintain historical information to enable the discovery of trends and developments over time. Hence data items usually contain time-related attributes like the time of a sales transaction or the order and shipping date of a product. Furthermore the values of these time-related attributes have a tendency to increase over time. We refer to this as the Multi-Append-Only-Trend (MAOT) property. In this paper we formalize the notion of MAOT and show how taking advantage of this property can improve query performance considerably. We focus on range aggregate queries which are essential for summarizing large data sets. Compared to MOLAP data cubes the amount of pre-computation and hence additional storage in the proposed technique is dramatically reduced. © Springer-Verlag Berlin Heidelberg 2003.