Exploiting the Multi-Append-Only-Trend Property of Historical Data in Data Warehouses

Data warehouses maintain historical information to enable thediscovery of trends and developments over time. Hence data itemsusually contain time-related attributes like the time of a salestransaction or the order and shipping date of a product.Furthermore the values of these time-related attributes have atendency to increase over time. We refer to this as theMulti-Append-Only-Trend (MAOT) property. In this paper weformalize the notion of MAOT and show how taking advantage of thisproperty can improve query performance considerably. We focus onrange aggregate queries which are essential for summarizing largedata sets. Compared to MOLAP data cubes the amount ofpre-computation and hence additional storage in the proposedtechnique is dramatically reduced.